CN115829900A - Detection method, detection system, device and storage medium - Google Patents

Abstract

A detection method, a detection system, a device and a storage medium, wherein the method comprises the following steps: acquiring an image to be processed with a plurality of pixel points; performing first identification processing on an image to be processed, and acquiring pixel points meeting a first threshold condition in the image to be processed as initial abnormal points; acquiring an abnormal point according to the initial abnormal point; selecting one or more pixel points adjacent to the abnormal point as peripheral pixel points; performing second identification processing on the peripheral pixel points, and taking the peripheral pixel points meeting a second threshold condition as defective pixel points, wherein the second threshold condition comprises a first threshold condition; and under the condition that the peripheral pixel points have the defective pixel points, judging that the abnormal points are defective points, and under the condition that the peripheral pixel points do not have the defective pixel points, judging that the abnormal points are noise points. And recognizing the peripheral pixel points by utilizing a second threshold condition, so that the peripheral pixel points can be detected under the condition that the difference between the peripheral pixel points and the normal pixel points is smaller, and thus, noise points and defect points are distinguished.

Description

Detection method, detection system, equipment and storage medium

technical field

The embodiments of the present invention relate to the technical field of optical detection, and in particular to a detection method, a detection system, a device, and a storage medium.

Background technique

With the continuous development of technology, precision machining is used in more and more fields, and at the same time, there are higher and higher requirements for machining accuracy. In order to meet the requirements of processing accuracy and improve the pass rate of products, it is necessary to carry out online inspection of products (for example, by performing defect detection to determine whether there are defects in the product, and to detect the position and size of defects), so as to ensure that the product manufacturing requirements are met. Relevant index requirements.

Among existing detection methods, optical detection is a general term for methods that utilize light to interact with analytes to achieve detection. Optical detection is not in contact with the object to be tested, has the characteristics of fast detection speed, no additional pollution, etc., and can realize online detection. Therefore, optical detection is widely used in the field of quality control of product manufacturing.

However, the accuracy of the current detection results still needs to be improved.

Contents of the invention

The problem to be solved by the embodiments of the present invention is to provide a light detection method, detection system, equipment and storage medium to improve the accuracy of detection results.

In order to solve the above problems, an embodiment of the present invention provides a detection method, including: acquiring an image to be processed, the image to be processed has a plurality of pixels; performing a first recognition process on the image to be processed, and acquiring the image to be processed Pixels satisfying the first threshold condition in the image are used as initial abnormal points; obtaining abnormal points according to the initial abnormal points; selecting one or more pixel points adjacent to the abnormal points as surrounding pixel points; for the surrounding pixels The second identification process is performed on the point, and the surrounding pixels satisfying the second threshold condition are regarded as defective pixels, and the second threshold condition includes the first threshold condition; in the case that there are defective pixels in the surrounding pixels , determining that the abnormal point is a defect point, and determining that the abnormal point is a noise point when there is no defective pixel point among the surrounding pixel points.

Correspondingly, an embodiment of the present invention also provides a detection system, including: an image acquisition module, configured to acquire an image to be processed, the image to be processed has a plurality of pixels; a first identification module, configured to identify the image to be processed The first recognition process is performed on the image, and the pixels satisfying the first threshold condition in the image to be processed are obtained as initial abnormal points; the abnormal point acquisition module is used to obtain abnormal points according to the initial abnormal points; the pixel point selection module uses Selecting one or more pixel points adjacent to the abnormal point as surrounding pixel points; the second identification module is used to perform a second identification process on the surrounding pixel points, and the surrounding pixel points satisfying the second threshold condition As a defective pixel, the second threshold condition includes the first threshold condition; a judging module, configured to determine that the abnormal point is a defective pixel when there is a defective pixel in the surrounding pixels, and in the If there is no defective pixel among the surrounding pixels, it is determined that the abnormal point is a noise point.

Correspondingly, an embodiment of the present invention also provides a device, including at least one memory and at least one processor, the memory stores one or more computer instructions, wherein the one or more computer instructions are executed by the processor Execute to realize the detection method described in the embodiment of the present invention.

Correspondingly, the embodiment of the present invention also provides a storage medium, the storage medium stores one or more computer instructions, and the one or more computer instructions are used to implement the detection method described in the embodiment of the present invention.

Compared with the prior art, the technical solutions of the embodiments of the present invention have the following advantages:

In the detection method provided by the embodiment of the present invention, the first recognition process is performed on the image to be processed, and after obtaining the pixel points satisfying the first threshold condition in the image to be processed as the initial abnormal point, the abnormal point is obtained according to the initial abnormal point, Selecting one or more pixels adjacent to the abnormal point as surrounding pixels, and performing a second identification process on the surrounding pixels, using surrounding pixels that meet the second threshold condition as defective pixels, the The second threshold condition includes the first threshold condition; wherein, the pixels adjacent to the noise point are normal pixels, and the pixels adjacent to the real defect are usually defective pixels, therefore, in the detection process, When an abnormal point appears, use the second threshold condition to identify the surrounding pixel points again. The second threshold condition includes the first threshold condition, that is, the set corresponding to the first threshold condition is included in the second threshold condition. In the set corresponding to the threshold condition, this makes it possible to detect when the difference between the surrounding pixels and the normal pixels is small, so that more defective pixels can be detected around the abnormal points, and the corresponding can be better. Distinguish between noise points and defect points to accurately judge whether the abnormal points are real defect points, so as to remove noise points and retain defect points, correspondingly reduce the probability of false detection and missed detection, and improve the accuracy of detection results.

In the detection system provided by the embodiment of the present invention, an abnormal point acquisition module, a pixel point selection module, a second identification module and a judgment module are set, and the first identification module performs the first identification processing on the image to be processed, and obtains the image to be processed After the pixel point meeting the first threshold condition is used as the initial abnormal point, the abnormal point acquisition module obtains the abnormal point according to the initial abnormal point, and the pixel point selection module is used to select one or more pixel points adjacent to the abnormal point as Surrounding pixels, and use the second identification module to perform a second identification process on the surrounding pixels, and use the surrounding pixels that meet the second threshold condition as defective pixels, and the second threshold condition includes the first threshold condition, the judging module judges whether the abnormal point is a defect or a noise point according to the result of the second identification module; wherein, the pixels adjacent to the noise point are all normal pixels, and the pixels adjacent to the real defect are usually also defects Therefore, in the detection process, when an abnormal point appears, the surrounding pixels are identified again by using the second threshold condition, and the second threshold condition includes the first threshold condition, that is, the first threshold The set corresponding to the condition is included in the set corresponding to the second threshold condition, which allows the surrounding pixels to be detected when the difference between the surrounding pixels and the normal pixels is small, so that more abnormal points can be detected around the abnormal point. More defect pixels can better distinguish noise points and defect points, so as to accurately judge whether the abnormal points are real defect points, which is convenient for removing noise points and retaining defect points, and correspondingly reduces the probability of false detection and missed detection , thereby improving the accuracy of the detection results.

Description of drawings

Fig. 1 is the flowchart of an embodiment of detection method of the present invention;

Fig. 2 is a partial schematic diagram of an image to be processed in an embodiment of step S1 in Fig. 1;

Fig. 3 is an enlarged view of an embodiment of any unit image in Fig. 2;

Fig. 4 is a schematic diagram of an embodiment of step S2 in Fig. 1;

Fig. 5 is a schematic diagram of an embodiment of the image to be tested in Fig. 4;

Fig. 6 is a schematic diagram of an embodiment of step S4 in Fig. 1;

Fig. 7 is a schematic diagram of an embodiment of step S5 in Fig. 1;

Fig. 8 is a functional block diagram of an embodiment of the detection system of the present invention;

FIG. 9 is a hardware structural diagram of a device provided by an embodiment of the present invention.

Detailed ways

It can be seen from the background technology that the accuracy of the current detection results still needs to be improved.

After research, it is found that in the actual detection process, the pixel size of small defects and noise detection is small (for example, usually one pixel), which makes it difficult to distinguish small defects and noise points. If both Eliminating it will miss a real defect, resulting in a missed detection, while keeping both will add a false detection.

In order to solve the technical problem, an embodiment of the present invention provides a detection method, including: acquiring an image to be processed, the image to be processed has a plurality of pixels; performing a first recognition process on the image to be processed, acquiring the The pixel points satisfying the first threshold condition in the image to be processed are used as initial abnormal points; the abnormal points are obtained according to the initial abnormal points; one or more pixel points adjacent to the abnormal points are selected as surrounding pixel points; The surrounding pixels are subjected to the second identification process, and the surrounding pixels satisfying the second threshold condition are regarded as defective pixels, and the second threshold condition includes the first threshold condition; among the surrounding pixels, there are defective pixels In some cases, it is determined that the abnormal point is a defect point, and in the case that there is no defective pixel point in the surrounding pixel points, it is determined that the abnormal point is a noise point.

The pixels adjacent to the noise point are normal pixels, and the pixels adjacent to the real defect are usually defective pixels. Therefore, in the detection process, when an abnormal point appears, the surrounding pixels are divided by the second threshold condition The point is identified again, the second threshold condition includes the first threshold condition, that is, the set corresponding to the first threshold condition is included in the set corresponding to the second threshold condition, which makes the surrounding pixel points and When the difference between normal pixels is small, they can still be detected, so that more defective pixels can be detected around the abnormal points, and correspondingly, noise points and defective points can be better distinguished to accurately judge the abnormal points Whether it is a real defect point is convenient for removing noise points and retaining defect points, correspondingly reducing the probability of false detection and missed detection, thereby improving the accuracy of detection results.

Referring to FIG. 1 , a flow chart of an embodiment of the detection method of the present invention is shown.

In this embodiment, the detection method includes the following basic steps:

Step S1: Acquiring an image to be processed, the image to be processed has a plurality of pixels;

Step S2: performing a first recognition process on the image to be processed, and acquiring pixels satisfying a first threshold condition in the image to be processed as initial abnormal points;

Step S3: obtaining abnormal points according to the initial abnormal points;

Step S4: selecting one or more pixel points adjacent to the abnormal point as surrounding pixel points;

Step S5: Perform a second identification process on the surrounding pixels, and use the surrounding pixels that meet the second threshold condition as defective pixels, and the second threshold condition includes the first threshold condition; If there is a defective pixel in the surrounding pixels, it is determined that the abnormal point is a defective point, and in the case that there is no defective pixel in the surrounding pixels, it is determined that the abnormal point is a noise point.

In order to make the above objects, features and advantages of the present invention more comprehensible, specific embodiments of the present invention will be described in detail below in conjunction with the accompanying drawings.

2 and FIG. 3 in combination, FIG. 2 is a partial schematic diagram of an embodiment of an image to be processed, and FIG. 3 is an enlarged view of an embodiment of any unit image in FIG. The image 100 to be processed has a plurality of pixels 110 (as shown in FIG. 2 ).

The image to be processed 100 is an image that needs to be detected. As an example, the image to be processed 100 is an image that requires defect detection.

In this embodiment, the image to be processed 100 is an image obtained by photographing the object to be tested. Specifically, the step of acquiring the image to be processed 100 includes: providing an imaging system and an object to be tested; using the imaging system, taking an image of the object to be tested as the image to be processed 100 .

In this embodiment, the object under test includes a plurality of repeated unit structures, therefore, the image to be processed 100 includes a plurality of identical unit images 150 . Wherein, the image to be processed 100 is an image of the plurality of unit structures, and the unit image 150 is an image of any unit structure.

It should be noted that, according to the arrangement of multiple repeated unit structures in the object under test, the unit images are arranged periodically. As shown in FIG. 2, as an example, only nine unit images 150 are shown. Specifically, the nine unit images 150 are arranged in a 3*3 array. It can be understood that the number of the unit images 150 is not limited to nine.

In this embodiment, the object under test is a wafer, and a wafer usually includes a plurality of repeated dies. Correspondingly, the image to be processed 100 is a wafer image, and each unit image 150 may include images of one die or multiple dies.

In other embodiments, the object under test may also be other types of products such as glass panels. It can be understood that the glass panel can also have multiple repeating unit structures. For example, each cell structure can be used to form a display screen of an electronic product.

It should also be noted that the smallest unit of an image is a pixel (pixel), therefore, the image 100 to be processed has a plurality of pixel points 110 . Specifically, the plurality of pixel points 110 constitute a pixel array.

In this embodiment, when the image to be processed 100 is subsequently detected, each of the unit images 150 is correspondingly detected.

4 and FIG. 5 in combination, FIG. 4 is a schematic diagram of an embodiment of step S2 in FIG. 1, and FIG. 5 is a schematic diagram of an embodiment of the image to be tested in FIG. In an identification process, pixel points 110 satisfying the first threshold condition in the image to be processed 100 are obtained as initial abnormal points 200 (as shown in FIG. 4 ).

Subsequently, the abnormal points are acquired according to the initial abnormal points 200 .

Correspondingly, the first identification process is used to prepare for the subsequent second identification process.

Here, the first threshold condition refers to: a set of pixel points 110 meeting the preset condition.

Correspondingly, the first threshold condition is used as a criterion for judging whether the pixel point 110 is the initial abnormal point 200, that is, if the pixel point 110 meets the first threshold value condition, then the pixel point 110 is the initial abnormal point 200 .

As an example, the first threshold condition is used to determine whether the intensity characteristic value of the pixel point 110 is abnormal.

In this embodiment, the intensity characteristic value is positively correlated with the gray value or the signal-to-noise ratio. Specifically, the intensity characteristic value includes a grayscale value or a light intensity value or a brightness value of the pixel point 110 forming the image 100 to be processed.

In this embodiment, the intensity characteristic value is described by taking the gray value as an example.

In this embodiment, the step of performing the first recognition process on the image to be processed 100 includes: acquiring a reference image; performing matching processing on the reference image and the image to be processed 100, so that the image to be processed is matched with the One-to-one correspondence between the pixels in the reference image matching area; compare the reference image with the image to be processed 100, and obtain the intensity characteristic value between the reference image and the corresponding pixel in the image to be processed 100 The first difference value; compare the first difference value with the first threshold value, and obtain the pixel points of the image to be processed whose first difference value is greater than the first threshold value as the first difference point; use the first difference point to obtain the initial 200 outliers.

The reference image is used as a comparison reference when performing recognition processing on the image to be processed 100, and the image to be tested is judged by comparing the first difference value between the intensity characteristic values of corresponding pixels in the image to be processed 100 and the reference image. Whether there is an initial outlier point 200 in 151.

In this embodiment, the first threshold condition at least includes: the first difference value is greater than the first threshold, therefore, acquire the pixels of the image to be processed whose first difference value is greater than the first threshold as the first difference point .

In this embodiment, in the step of performing the first identification process on the image to be processed 100, the first identification process is performed on each of the unit images 150, the initial abnormal point 200 of each unit image 150 is obtained, and the current The unit image 150 to be tested is used as an image to be tested 151 , and multiple unit images around the image to be tested 151 are used as reference images 152 .

Correspondingly, a matching process is performed on the reference image 152 and the image to be tested 151, and after the matching process, the reference image 152 is compared with the image to be tested 151, and the reference image 152 and the image to be tested are obtained. The first difference value between the intensity characteristic values of corresponding pixel points in the image-to-be-tested 151.

Here, the corresponding pixel points in the image-to-be-test 151 and the reference image 152 refer to: the pixel points at the same position in the image-to-be-test 151 and the reference image 152 .

The reference image 152 is used as a reference for comparison when performing recognition processing on the image to be tested 151, and by comparing the difference between the intensity characteristic values of corresponding pixels in the image to be tested 151 and the reference image 152, it is judged that Check whether there is an abnormal point 200 in the image 151.

The reference image 152 and the image to be tested 151 are located in the same image to be processed 100, that is to say, the reference image 152 and the image to be tested 151 come from the same object to be tested, thereby avoiding differences between different objects to be tested. The problem of large differences in the average intensity characterization value will have a negative impact on the accuracy of the detection results, which is conducive to improving the accuracy of the detection results.

Specifically, a plurality of unit images adjacent to the image-to-be-tested 151 serve as reference images 152 . As shown in FIG. 4 , as an example, the number of the reference images 152 is 8, and the reference images 152 and the images to be tested 151 are arranged in a 3*3 array.

In other embodiments, a standard image may also be selected as the reference image. Correspondingly, by using a standard image-based matching method, the pixels in the matching area of the image to be processed and the reference image are in one-to-one correspondence.

The standard image is an image obtained by shooting a reference object that is consistent with the object to be measured. It can be understood that the standard image and the image to be processed are not located on the same image. Wherein, the standard image may include a CAD drawing or a defect-free measurement image.

When a standard image is used as a reference image, the method for matching the reference image with the image to be processed includes a step of obtaining a matching area, and the step of obtaining a matching area includes: obtaining the highest matching degree between the image to be processed and the reference image or an area larger than a preset value to obtain a matching area, and each pixel point of the reference image has a one-to-one correspondence relationship with each pixel point of the matching area.

Specifically, the highest matching degree refers to: the variance, standard deviation or absolute value of the intensity characteristic value between each pixel point of the matching area and each pixel point of the reference image is the smallest; the matching degree is greater than the preset value means It is: the variance, standard deviation or absolute value of the intensity characteristic value between each pixel point of the matching area and each pixel point of the reference image is smaller than a preset value.

Wherein, the difference between the intensity characteristic value of the pixel in the matching area and the corresponding pixel in the reference image is a pixel difference, and the variance of the intensity characteristic value refers to: the sum of the squares of the pixel difference.

In this embodiment, the image to be processed includes a plurality of identical unit images, and the reference image is a standard image of a unit image. Therefore, matching processing is performed on the reference image and the image to be processed so that the The step of one-to-one correspondence between the image to be processed and the pixel points of the matching area of the reference image further includes: repeating the step of obtaining the matching area, obtaining a plurality of matching areas in the image to be processed, the number of the matching areas is the same as The number of the unit images is the same.

Taking a non-defective measurement image as an example as a standard image, by selecting the same reference object as the object to be tested, the reference object also has a plurality of unit structures, acquiring the image of the reference object, and A qualified unit structure image is selected from the image of the reference object as a standard image (that is, a reference image). For example, an image of a qualified wafer is selected first, and part of the images of qualified grains are selected on the image of the selected wafer as a reference image.

Correspondingly, if a standard image is selected as the reference image, each unit image is compared with the reference image when the recognition process is performed.

In this embodiment, the step of using the first difference point to obtain the initial abnormal point 200 includes: obtaining the number of the first difference point to obtain the first difference number, and when the first difference number is greater than or equal to the preset number In this case, it is determined that the currently detected pixel point is the initial abnormal point 200 .

Correspondingly, the first threshold condition further includes: the first difference quantity is greater than or equal to a preset quantity.

It should be noted that, when the first difference number is greater than or equal to the preset number, it is determined that the currently detected pixel point is the initial abnormal point 200, where the minimum value of the preset number is one, and the maximum value is The total number of reference images 152 corresponding to the measured image 151, the preset number can be set according to actual needs.

It should also be noted that, in other embodiments, the first difference point may also be directly used as the initial abnormal point 200 .

When comparing the image to be tested 151 with the reference image 152, the grayscale value of the pixel 110 of the image to be tested 151 is compared with the grayscale value of the corresponding pixel of the reference image 152, thereby obtaining the The first difference value between the gray levels of corresponding pixels of the image-to-be-tested 151 and the reference image 152 .

Here, the first difference value between the gray levels refers to the absolute value of the difference between the gray levels of the pixels corresponding to the image-to-be-tested 151 and the reference image 152 .

In this embodiment, the first threshold is positively correlated with the definition of the image 100 to be processed. The higher the definition of the image, the more obvious the grayscale change of the contour edge, and the stronger the sense of hierarchy. Therefore, in order to filter out the abnormal points 200, the value of the first threshold is correspondingly larger.

Specifically, the sharpness of the image is related to the grayscale gradient, therefore, the first threshold is obtained through the grayscale gradient.

It should be noted that, subsequently, the second identification process is performed on the surrounding pixels to determine whether the surrounding pixels meet the second threshold condition, and the surrounding pixels satisfying the second threshold condition are regarded as defective pixels, and the The second threshold condition includes the first threshold condition, therefore, the second threshold condition is correlated with the first threshold condition.

Therefore, the step of obtaining the first threshold and/or obtaining the second threshold includes an initial threshold obtaining process, and the initial threshold obtaining process includes: obtaining the gray gradient of each pixel point 110 in the image to be processed 100; The average value of the grayscale gradient of the image to be processed 100 is acquired, and the average value of the grayscale gradient is used as an initial threshold.

In this embodiment, the detection method further includes: adding a preset offset based on the initial threshold. It should be noted that the preset offset here may or may not be zero.

In this embodiment, a preset offset is added based on the initial threshold to obtain the first threshold and/or the second threshold.

In this embodiment, the first threshold can be obtained first according to the initial threshold, and the second threshold can be obtained by using the first threshold, or the second threshold can be obtained first according to the initial threshold, and the first threshold can be obtained by using the second threshold. Threshold, get the first threshold and the second threshold.

Specifically, the step of obtaining the first threshold further includes: using the initial threshold as the first threshold; the step of obtaining the second threshold correspondingly includes: multiplying the first threshold by a first scaling factor to obtain the second threshold; or, the step of obtaining the second threshold further includes: using the initial threshold as the second threshold; the step of obtaining the first threshold correspondingly includes: making the second threshold and the second ratio multiplied by factors to obtain the first threshold; or, the steps of obtaining the first threshold and obtaining the second threshold both include an initial threshold obtaining process, and the preset offset corresponding to the first threshold is the same as the The preset offsets corresponding to the second thresholds are different.

It can be understood that, when the step of obtaining the first threshold further includes using the initial threshold as the first threshold, the preset offset corresponding to the first threshold is correspondingly zero.

Similarly, if the step of obtaining the second threshold further includes using the initial threshold as the second threshold, the preset offset corresponding to the second threshold is correspondingly zero.

In this embodiment, the first threshold is obtained by adding a preset offset based on the initial threshold.

It should be noted that any pixel point 110 has a corresponding grayscale gradient in the X direction and Y direction, therefore, as an example, the relational expression of the grayscale gradient of the pixel point includes:

Wherein, M(x, y) represents the gray gradient of the pixel point 110, gx represents the gradient of the pixel point 110 in the X direction, gy represents the gradient of the pixel point 110 in the Y direction, and the X The direction is the row direction of the pixel array, and the Y direction is the column direction of the pixel array.

It should also be noted that, in the step of obtaining the first threshold by adding a preset offset based on the initial threshold, the preset offset should not be too small or too large. If the preset offset is too small, it is easy to cause the false detection rate to become high when performing the first identification process, that is, it is easy to classify the normal pixel point 110 as the initial abnormal point 200, resulting in the initial abnormal point 200 If the number is too large, the amount of subsequent data processing for obtaining abnormal points according to the initial abnormal points and the second identification process will be increased; if the preset offset is too large, the probability of missed detection will easily increase. Therefore, in this embodiment, the preset offset is 3-5.

In addition, only one pixel point 110 is shown as the initial abnormal point 200 in FIG. 5 , but the initial abnormal point 200 is not limited to one pixel point. For example, in other embodiments, after the first identification process, multiple initial abnormal points are obtained, and there is a situation that multiple initial abnormal points are connected.

Continuing to refer to FIG. 5 , step S3 is executed to obtain the abnormal point 250 according to the initial abnormal point 200 .

Subsequently, by further identifying the abnormal point 250, it is judged whether the abnormal point 250 is a defect point or a noise point.

In this embodiment, obtaining the abnormal point 250 according to the initial abnormal point 200 includes: performing connected domain judgment on the initial abnormal point 200 in the image 100 to be processed, obtaining a connected component (Connected Component), and each connected component in the same connected domain Each initial outlier point 200 has adjacent initial outlier points 200, and the initial outlier points 200 in the connected domain and outside the connected domain are separated from each other; in the case where the number of initial outlier points 200 in the connected domain is greater than or equal to the preset number , the initial outlier points 200 in the connected domain are used as defect points, and when the number of initial outlier points 200 in the connected domain is less than a preset number, the connected domain is used as the outlier points 250 .

A collection of multiple connected pixel points 110 meeting certain conditions is called a connected domain. Specifically in this embodiment, the specific condition is to meet the first threshold condition.

The outlier point 250 is rechecked in the second identification process. Therefore, it is determined whether the second identification process needs to be performed by first performing connected domain judgment.

Specifically, in the case where it is difficult to determine whether the type of the initial outlier 200 is a defect point or an outlier 250, the connected domain is judged first, so that the number of initial outlier points 200 in the connected domain is greater than or equal to the preset number In the case of , it can be determined that the initial abnormal point 200 is a defect point, and then there is no need to select surrounding pixel points and perform the second identification process, so that the number of initial abnormal points 200 in the connected domain can be selectively less than In the case of a preset number, the abnormal points 250 are re-inspected, thereby improving the detection efficiency while improving the accuracy of the detection results.

In this embodiment, the determination of connected domains includes determination of four connected domains or determination of eight connected domains. Among them, the four-connected domain judgment refers to: judging whether there are 4 adjacent initial abnormal points 200 around any initial abnormal point 200; the eight-connected domain judgment refers to: judging whether there are 8 adjacent initial abnormal points 200 Adjacent initial outliers 200.

When the number of 200 initial abnormal points in the connected domain is greater than or equal to the preset number, the initial abnormal points 200 in the connected domain are used as defect points, and the number of 200 initial abnormal points in the connected domain is less than the preset number In the case of a large number, the connected domain is used as the abnormal point 250, and the pixel size of the noise point is usually small. Therefore, if the preset number is too large, it is easy to misjudge the defective point as the abnormal point when judging the connected domain. Abnormal point 250, so unnecessary follow-up operations are performed. Therefore, in this embodiment, the preset number is 1 to 5.

As shown in FIG. 5 , as an example, the preset number is one. Correspondingly, if the initial outlier point 200 is an isolated pixel point 110, the initial outlier point 200 is taken as the outlier point 250; otherwise, the initial outlier point 200 is taken as a defect point. The pixel size of the noise point is generally small, usually one pixel point, therefore, by setting the preset number to one, it is beneficial to accurately obtain the abnormal point 250 and reduce the computation load of subsequent operations.

It should be noted that, when the number of initial outliers in the connected domain is less than the preset number, using the connected domain as the outlier refers to taking all the initial outliers in the connected domain as a Outlier.

In other embodiments, obtaining the abnormal point according to the initial abnormal point includes: obtaining pixel points adjacent to the initial abnormal point as adjacent points, and the number of the adjacent points is one or more; judging each adjacent point Whether it is an initial outlier point, and if none of the adjacent points is an initial outlier point, use the initial outlier point as an outlier point.

Specifically, the number of the adjacent points may be four or eight. When the number of the adjacent points is four, the initial abnormal point and the adjacent points are in a cross-shaped layout or an X-shaped layout; when the number of the adjacent points is eight, the initial abnormal point The abnormal points and the adjacent points are arranged in a 3*3 array.

Referring to FIG. 6, FIG. 6 is a schematic diagram of an embodiment of step S4 in FIG.

Subsequently, the second identification process is performed on the surrounding pixel points 300, so as to realize the re-examination of the abnormal point 250.

The pixel size of noise (noise) is usually small (for example, the pixel size of one noise is one pixel), which makes it difficult to distinguish defects from noise. However, the pixels adjacent to the noise point are all normal pixels, and the pixels adjacent to the real defect are usually also defective pixels. Therefore, in the detection process, when abnormal points 250 appear, the second threshold condition is used to The surrounding pixel points 300 are identified again, the second threshold condition includes the first threshold condition, that is, the set corresponding to the first threshold condition is included in the set corresponding to the second threshold condition, which makes In the case where the difference between the surrounding pixel 300 and the normal pixel is small, it can still be detected, so that more defective pixels can be detected around the abnormal point 250, and correspondingly, noise points and defective points can be better distinguished. By accurately judging whether the abnormal point 250 is a real defect point, it is convenient to eliminate noise points and retain the defect points, correspondingly reducing the probability of false detection and missed detection, thereby improving the accuracy of detection results.

Among them, the defect point represents the pixel point 110 corresponding to the real defect in the object under test forming the image 100 to be processed, and the noise point is also called noise. Anomalies, such as electronic disturbances in detectors, occur.

In this embodiment, by selecting the pixel point 110 next to the abnormal point 250 as the surrounding pixel point 300, the correlation between the detection result of the surrounding pixel point 300 and the abnormal point 250 is higher, thereby improving the subsequent second identification. Accuracy of processed test results.

In this embodiment, in the step of selecting one or more pixel points 110 adjacent to the abnormal point 250 as the surrounding pixel points 300, select the pixel points adjacent to the edge and/or vertex corner of the abnormal point 250 110 as surrounding pixel 300. Specifically, the number of the surrounding pixel points 300 is 8, and the surrounding pixel points 300 and the abnormal point 250 form a 3*3 pixel array. That is to say, the abnormal point 250 is the central pixel point in the 3*3 pixel array, and the surrounding pixel point 300 is the pixel point 110 next to the abnormal point 250 .

The abnormal point 250 is the central pixel point in the 3*3 pixel array, therefore, by selecting 8 surrounding pixel points 300, it is beneficial to improve the result accuracy of the second recognition process.

In other embodiments, the number of surrounding pixel points is four, and the surrounding pixel points and the abnormal point are in a cross-shaped layout or an X-shaped layout.

It should be noted that, in the process of obtaining outliers 250 according to the initial outliers 200, if the number of initial outliers 200 in the connected domain is less than the preset number, the connected domain is used as the outlier 250 . Wherein, when the preset number is multiple, using the connected domain as the outlier point 250 refers to: taking all initial outlier points 200 in the connected domain as one outlier point 250 . Select the pixel point 110 adjacent to the edge and/or vertex of the abnormal point 250 as the surrounding pixel point 300, then the number of surrounding pixel points 300 of the abnormal point 250 can be greater than 8, for example, when the abnormal point 250 When two initial abnormal points 200 are included, the number of surrounding pixel points 300 is 10 or 6.

It should also be noted that, in other embodiments, when the abnormal point is obtained by obtaining the pixel points adjacent to the initial abnormal point as adjacent points, one or more pixel points adjacent to the abnormal point are selected as The surrounding pixel points include: the neighboring pixel points of the initial abnormal point as the surrounding pixel points.

Referring to FIG. 7 in conjunction with FIG. 7, FIG. 7 is a schematic diagram of an embodiment of step S5 in FIG. A schematic diagram of the case where there are no defective pixels in the surrounding pixels 300, step S4 is executed, the second identification process is performed on the surrounding pixels 300, and the surrounding pixels 300 satisfying the second threshold condition are regarded as defective pixels , the second threshold condition includes the first threshold condition; if there is a defective pixel in the surrounding pixel 300, it is determined that the abnormal point 200 is a defective point, and there is no defective pixel in the surrounding pixel 300 If there are defective pixels, it is determined that the abnormal point 200 is a noise point.

Here, the second threshold condition refers to: a set of surrounding pixel points 300 meeting the preset condition.

The second threshold condition is used as a criterion for judging whether the surrounding pixel 300 is a defective pixel, that is, if the surrounding pixel 300 meets the second threshold condition, the surrounding pixel 300 is regarded as a defective pixel point.

In this embodiment, the second threshold condition includes the first threshold condition, that is, the set corresponding to the first threshold condition is included in the set corresponding to the second threshold condition, so that the second identification process detects The number of pixels that are different from normal pixels is increased.

If the abnormal point 200 is a defect point, there is usually a difference between the surrounding pixel points 300 around the abnormal point 200 and the normal pixel points 110, but the difference is small, and the pixels around the noise point are all Normal pixels, therefore, use the second threshold condition to carry out the second identification process, if the abnormal point 200 is a defect point, as shown in Figure 7 (a), after the second identification process, it can also be in the One or more defective pixel points are detected around the abnormal point 200. For example, FIG. If the abnormal point 200 is a noise point, as shown in FIG. 7(b), after the second identification process, no defective pixel points can be detected around the abnormal point 200, so that the true value of the abnormal point 200 Types are distinguished, that is, noise points and defect points are distinguished, so that the noise points can be removed later and the defect points can be retained.

In this embodiment, the second identification process is performed in the same way as the first identification process, and the threshold condition is changed.

In this embodiment, the step of performing the second recognition processing on the surrounding pixel points 300 includes: acquiring a reference image; performing matching processing on the reference image and the image to be processed 100, so that the image to be processed is matched with the One-to-one correspondence between the pixels in the matching area of the reference image; comparing the surrounding pixels 300 with the corresponding pixels in the reference image, and obtaining the intensity of the surrounding pixels 300 and the corresponding pixels in the reference image A second difference value between the characteristic values; comparing the second difference value with a second threshold value, and obtaining the surrounding pixel points 300 whose second difference value is greater than the second threshold value as a second difference point; using the second difference value Points to get defective pixels.

Correspondingly, in this embodiment, the second threshold condition at least includes: the second difference value is greater than a second threshold, wherein the second threshold is smaller than the first threshold.

In this embodiment, when performing the first identification process, the first difference value between the intensity characteristic value of the corresponding pixel point in the reference image and the image to be processed 100 is obtained, and the first difference value is compared and the first threshold, obtain the pixel points of the image to be processed whose first difference value is greater than the first threshold as the first difference point, and obtain the difference between the surrounding pixel points 300 and the reference image when performing the second identification process. A second difference value between the intensity characteristic values of corresponding pixels, and comparing the second difference value with a second threshold value, and acquiring the surrounding pixel points 300 whose second difference value is greater than the second threshold value as the second difference point , therefore, by making the second threshold smaller than the first threshold, the second threshold condition includes the first threshold condition.

Specifically, the step of using the second difference points to obtain defective pixels includes: obtaining the number of the second difference points to obtain a second difference number, and when the second difference number is greater than or equal to the preset number Next, it is determined that the currently detected peripheral pixel point 300 is a defective pixel point.

Correspondingly, the second threshold condition further includes: the second difference quantity is greater than or equal to a preset quantity.

Wherein, in the case where the second difference number is greater than or equal to the preset number, it is determined that the currently detected peripheral pixel point 300 is a defective pixel point, where the minimum value of the preset number is one, and the maximum value is one currently to be tested. The total number of reference images 152 corresponding to the surrounding pixel points 300, the preset number can be set according to actual needs.

It should be noted that, in other embodiments, the second difference point may also be directly used as a defective pixel point.

It should also be noted that the image to be tested 151 in the second recognition process is the unit image 150 where the abnormal point 200 is located.

In this embodiment, the second threshold is positively correlated with the definition of the image 100 to be processed.

In this embodiment, the steps of obtaining the first threshold and obtaining the second threshold both include the step of obtaining an initial threshold, and the preset offset corresponding to the first threshold is equal to the preset offset corresponding to the second threshold. The displacement is not the same.

When the second threshold is smaller than the first threshold, after the second threshold is obtained, the ratio of the second threshold to the first threshold should not be too small, nor should it be too large. If the ratio of the second threshold to the first threshold is too small, it is easy to cause the second threshold to be too small, and when the second identification process is performed, it is easy to cause the false detection rate to become high, that is to say, it is easy to Classify the normal surrounding pixels 300 as defective pixels, so that it is easy to mistakenly classify the abnormal points 200 as defective points; if the ratio of the second threshold to the first threshold is too large, it is easy to increase the probability of missed detection , so it is easy to misclassify the abnormal point 200 as a noise point. Therefore, in this embodiment, the second threshold is 60% to 80% of the first threshold.

It should be noted that, in other embodiments, the first identification process and the second identification process may also be performed in other manners.

Specifically, in some other embodiments, when the image to be processed is a dark field image, the step of performing the first recognition process on the image to be processed includes: combining the pixels of the image to be processed with the first A threshold value is compared, and the pixel point whose intensity characteristic value is greater than the first threshold value is acquired as an initial abnormal point; the first threshold value condition includes: the intensity characteristic value of the pixel point is greater than the first threshold value.

Correspondingly, the step of performing the second identification process on the surrounding pixels includes: comparing the surrounding pixels with a second threshold, and obtaining the surrounding pixels whose intensity characteristic value is greater than the second threshold as defects A pixel point; the second threshold condition includes: the intensity characteristic value of the surrounding pixel point is greater than the second threshold; wherein, the second threshold is smaller than the first threshold.

The dark field image is an image obtained by means of dark field inspection. In the dark field image, the grayscale of the background is smaller than the grayscale of the defect image on the surface of the object to be tested.

In optical inspection, according to the source of collected signal light, bright field inspection (brightfield inspection) and dark field inspection can be included. Among them, dark field detection is a method of detecting the surface of the object to be measured by detecting the intensity of scattered light on the surface of the object to be measured.

In other embodiments, when the image to be processed is a bright field image, the step of performing the first identification process on the image to be processed includes: performing a first threshold value on each pixel of the image to be processed In comparison, acquiring the pixel whose intensity characteristic value is less than the first threshold is taken as an initial abnormal point; the first threshold condition includes: the intensity characteristic value of the pixel is less than the first threshold.

Correspondingly, the step of performing the second identification process on the surrounding pixels includes: comparing the surrounding pixels with a second threshold, and acquiring the surrounding pixels whose intensity characteristic value is smaller than the second threshold as defects A pixel point; the second threshold condition includes: the intensity characteristic value of the surrounding pixel point is smaller than the second threshold; wherein, the second threshold is greater than the first threshold.

During the first identification process, acquire the pixel points whose intensity characteristic value is greater than the first threshold value as the initial abnormal point; during the second identification process, acquire the surrounding area whose intensity characteristic value is greater than the second threshold value A pixel is a defective pixel, therefore, by making the second threshold smaller than the first threshold, the second threshold condition includes the first threshold condition.

The bright field image is an image obtained by means of bright field inspection. In the bright field image, the grayscale of the background is greater than the grayscale of the defect image on the surface of the object to be tested.

Among them, bright field detection is a method of detecting the surface of the object to be measured by detecting the intensity of reflected light on the surface of the object to be measured.

It should also be noted that if the second threshold condition is directly used for the first recognition process, it will easily lead to a high false detection rate (for example, the pixel with a certain intensity characteristic value difference, but the intensity characteristic value difference is within an acceptable range points are classified as defective points), therefore, in this embodiment, the first recognition process is first performed using the first threshold condition to obtain initial abnormal points, and obtain abnormal points according to the initial abnormal points to exclude normal pixel points 110 or The pixel points 110 whose intensity characteristic value difference is within the acceptable range or the pixel points 110 determined to be defective points, so that the number of abnormal points 250 will not be too many, and then use the second threshold value condition to perform the second identification process on the surrounding pixel points 300 , so as to realize the re-inspection of the abnormal point 250, so that the real defect point can be screened out more accurately while reducing the data processing amount of the second identification process.

Correspondingly, the embodiment of the present invention also provides a detection system. Referring to FIG. 8 , it shows a functional block diagram of an embodiment of the detection system of the present invention.

With reference to FIGS. 2 to 7, the detection system includes: an image acquisition module 10, configured to acquire an image to be processed 100, the image to be processed 100 having a plurality of pixels 110; a first recognition module 20, configured to identify the The image to be processed 100 is subjected to the first identification process, and the pixel point 110 satisfying the first threshold condition in the image to be processed is obtained as the initial abnormal point 200; the abnormal point obtaining module 70 is used to obtain the abnormal point according to the initial abnormal point 200 Point 250; pixel point selection module 30, used to select one or more pixel points 110 adjacent to the abnormal point 200 as surrounding pixel points 300; second identification module 40, used to perform on the surrounding pixel points 300 The second identification process is to use the surrounding pixel points 300 satisfying the second threshold condition as defective pixel points, and the second threshold condition includes the first threshold condition; the judging module 50 is used to identify the surrounding pixel points 300 If there is a defective pixel point in the surrounding pixel point 300, it is determined that the abnormal point 200 is a defect point, and if there is no defective pixel point in the surrounding pixel point 300, it is determined that the abnormal point 200 is a noise point.

The pixels adjacent to the noise point are all normal pixels, and the pixels adjacent to the real defect are usually defective pixels. Therefore, in the detection process, when an abnormal point appears, the second threshold condition is used to convert the The surrounding pixels are identified again, the second threshold condition includes the first threshold condition, that is, the set corresponding to the first threshold condition is included in the set corresponding to the second threshold condition, which makes the surrounding pixels When the difference between normal pixels and normal pixels is small, it can still be detected, so that more defective pixels can be detected around abnormal points, and correspondingly, noise points and defective points can be better distinguished, so as to accurately judge the Whether the abnormal point is a real defect point is convenient for removing noise points and retaining defect points, correspondingly reducing the probability of false detection and missed detection, thereby improving the accuracy of detection results.

The image acquisition module 10 is used to acquire an image to be processed 100, and the image to be processed 100 is an image to be detected. As an example, the image to be processed 100 is an image that requires defect detection.

In this embodiment, the image to be processed 100 is an image obtained by photographing the object to be tested.

Specifically, the image acquisition module 10 includes an imaging system, and the imaging system is used to capture and acquire an image of the object under test as an image to be processed 100 .

The object under test includes a plurality of repeated unit structures, therefore, the image to be processed 100 includes a plurality of identical unit images 150 . Wherein, the image to be processed 100 is an image of the plurality of unit structures, and the unit image 150 is an image of any unit structure.

It should be noted that, according to the arrangement of multiple repeated unit structures in the object under test, the unit images are arranged periodically.

As shown in FIG. 2, as an example, only nine unit images 150 are shown. Specifically, the nine unit images 150 are arranged in a 3*3 array. It can be understood that the number of the unit images 150 is not limited to nine.

Correspondingly, the detection object of the detection system is each unit image 150 in the image to be processed 100 .

In this embodiment, the object under test is a wafer, and a wafer usually includes a plurality of repeated dies. Correspondingly, the image to be processed 100 is a wafer image, and each unit image 150 may include images of one die or multiple dies.

In other embodiments, the object under test may also be other types of products such as glass panels. It can be understood that the glass panel can also have multiple repeating unit structures. For example, each cell structure can be used to form a display screen of an electronic product.

It should also be noted that the smallest unit of an image is a pixel (pixel), therefore, the image 100 to be processed has a plurality of pixel points 110 . Specifically, the plurality of pixel points 110 constitute a pixel array.

The first identification module 20 is configured to perform a first identification process on the image to be processed 100 to obtain an initial abnormal point 200, so as to prepare for subsequent acquisition of abnormal points.

Specifically, the first recognition module 20 uses a first threshold condition to perform a first recognition process on the image to be processed 100 . Here, the first threshold condition refers to: a set of pixel points 110 meeting the preset condition. The first threshold condition is used as a criterion for judging whether the pixel point 110 is the initial abnormal point 200 , that is, if the pixel point 110 meets the first threshold value condition, the pixel point 110 is the initial abnormal point 200 .

As an example, the first threshold condition is used to determine whether the intensity characteristic value of the pixel point 110 is abnormal.

In this embodiment, the intensity characteristic value is positively correlated with the gray value or the signal-to-noise ratio. Specifically, the intensity characteristic value includes a grayscale value or a light intensity value or a brightness value of the pixel point 110 forming the image 100 to be processed.

In this embodiment, the intensity characteristic value is described by taking the gray value as an example.

In this embodiment, the first identification module 20 includes: a first threshold setting unit, configured to set a first threshold condition; a first image selection unit, configured to obtain an image to be processed 100; a second image selection unit, configured to Acquiring a reference image; a first matching unit, configured to perform matching processing on the reference image and the image to be processed 100, so that the pixels of the image to be processed and the matching area of the reference image correspond one-to-one; the first comparison A unit, configured to compare the reference image with the image to be processed 100, and obtain a first difference value between intensity representation values of corresponding pixels in the reference image and the image to be processed 100; second A comparison unit, configured to compare the first difference value with a first threshold, and acquire pixels of the image to be processed whose first difference value is greater than the first threshold as a first difference point; a first abnormal point acquisition unit, configured to The initial outlier point 200 is obtained by using the first difference point.

The reference image is used as a reference for comparison when performing recognition processing on the pixel points 110 in the image to be processed 100, by comparing the first difference value between the intensity characteristic values of the corresponding pixel points in the image to be processed 100 and the reference image, Therefore, it is determined whether there is an initial abnormal point 200 in the image to be tested 151 .

In this embodiment, the first threshold condition at least includes: the first difference value is greater than the first threshold, therefore, acquire the pixels of the image to be processed whose first difference value is greater than the first threshold as the first difference point .

In this embodiment, the first recognition module 20 is used to perform the first recognition process on each of the unit images 150, acquire the initial abnormal point 200 of each unit image 150, and compare the unit images to be tested 150 is used as an image to be tested 151 , and a plurality of unit images around the image to be tested 151 are used as reference images 152 .

Correspondingly, the first matching unit is configured to perform matching processing on the reference image 152 and the image-to-be-tested 151; the first comparison unit is configured to match the reference image 152 to the image-to-be-tested 151 to compare and obtain a first difference value between the intensity characteristic values of corresponding pixels in the reference image 152 and the image-to-be-tested 151 .

Here, the corresponding pixel points in the image-to-be-test 151 and the reference image 152 refer to: the pixel points at the same position in the image-to-be-test 151 and the reference image 152 .

The reference image 152 is used as a reference for comparison when performing recognition processing on the image to be tested 151, and by comparing the difference between the intensity characteristic values of corresponding pixels in the image to be tested 151 and the reference image 152, it is judged that Check whether there is an abnormal point 200 in the image 151.

The reference image 152 and the image to be tested 151 are located in the same image to be processed 100, that is, the reference image 152 and the image to be tested 151 are from the same object to be tested, thereby avoiding the average The problem of large differences in intensity characterization values will have a negative impact on the accuracy of the test results, and it will be beneficial to improve the accuracy of the test results.

Specifically, a plurality of unit images adjacent to the image-to-be-tested 151 serve as reference images 152 . As shown in FIG. 4 , as an example, the number of the reference images 152 is 8, and the reference images 152 and the images to be tested 151 are arranged in a 3*3 array.

In other embodiments, the second image selection unit may also select a standard image as the reference image. Correspondingly, the first matching unit makes a one-to-one correspondence between the image to be processed and the pixels in the matching area of the reference image by means of matching based on a standard image.

The standard image is an image obtained by shooting a reference object that is consistent with the object to be measured. It can be understood that the standard image and the image to be processed are not located on the same image. Wherein, the standard image may include a CAD drawing or a defect-free measurement image.

When a standard image is used as a reference image, the first matching unit is configured to perform a step of acquiring a matching area. Specifically, the first matching unit is used to obtain the area in the image to be processed that has the highest matching degree with the reference image or greater than a preset value to obtain the matching area, and each pixel point of the reference image and the pixel point of the matching area have a one-to-one relationship.

Specifically, the highest matching degree refers to: the variance, standard deviation or absolute value of the intensity characteristic value between each pixel point of the matching area and each pixel point of the reference image is the smallest; the matching degree is greater than the preset value means It is: the variance, standard deviation or absolute value of the intensity characteristic value between each pixel point of the matching area and each pixel point of the reference image is smaller than a preset value. Wherein, the difference between the intensity characteristic value of the pixel in the matching area and the corresponding pixel in the reference image is a pixel difference, and the variance of the intensity characteristic value refers to: the sum of the squares of the pixel difference.

In this embodiment, the image to be processed includes a plurality of identical unit images, and the reference image is a standard image of a unit image, therefore, the first matching unit is used to repeat the step of obtaining the matching area, and in the Acquiring multiple matching regions from the image to be processed, the number of the matching regions is the same as the number of the unit images.

For example, selecting the same reference object as the object to be tested, the reference object also has a plurality of unit structures, the second image selection unit acquires the image of the reference object, and selects the reference object on the image of the reference object Qualified cell structure image, as a reference image. For example, an image of a qualified wafer is selected first, and part of the images of qualified grains are selected on the image of the selected wafer as a reference image.

Correspondingly, if a standard image is selected as the reference image, the first comparison unit is used to compare each unit image with the reference image.

In this embodiment, the first abnormal point acquisition unit includes: a first difference quantity acquisition subunit, configured to acquire the number of the first difference points, to obtain the first difference quantity; a first screening subunit, configured to When the first difference amount is greater than or equal to the preset amount, it is determined that the currently detected pixel point is the initial abnormal point 200.

Correspondingly, the first threshold condition further includes: the first difference quantity is greater than or equal to a preset quantity.

It should be noted that the minimum value of the preset number here is one, and the maximum value is the total number of reference images 152 corresponding to the image-to-be-tested 151 , and the preset number can be set according to actual needs.

It should also be noted that, in other embodiments, the first abnormal point obtaining unit may also directly use the first difference point as an initial abnormal point.

When the first comparing unit compares the image-to-be-tested 151 with the reference image 152, it compares the grayscale value of the pixel 110 of the image-to-be-tested 151 with the grayscale value of the corresponding pixel of the reference image 152 , so as to obtain the first difference value between the gray levels of corresponding pixels in the image-to-be-test 151 and the reference image 152 .

Here, the first difference value between the gray levels refers to the absolute value of the difference between the gray levels of the pixels corresponding to the image-to-be-tested 151 and the reference image 152 .

In this embodiment, the first threshold is positively correlated with the definition of the image 100 to be processed. The higher the definition of the image, the more obvious the grayscale change of the contour edge, and the stronger the sense of hierarchy. Therefore, in order to filter out the abnormal points 200, the value of the first threshold is correspondingly larger.

Specifically, the sharpness of the image is related to the grayscale gradient, therefore, the first threshold setting unit acquires the first threshold through the grayscale gradient.

It should be noted that, subsequently, the second identification process is performed on the surrounding pixels to determine whether the surrounding pixels meet the second threshold condition, and the surrounding pixels satisfying the second threshold condition are regarded as defective pixels, and the The second threshold condition includes the first threshold condition, therefore, the second threshold condition is correlated with the first threshold condition.

In this embodiment, the first threshold and the second threshold are obtained through the initial threshold.

Therefore, the detection system further includes: an initial threshold acquisition module 60, configured to acquire an initial threshold.

In this embodiment, the initial threshold acquisition module 60 includes: a first gradient acquisition unit, configured to acquire the gray gradient of each pixel point 110 in the image to be processed 100; a second gradient acquisition unit, configured to acquire the The average value of the gray gradient of the image 100 to be processed is used as the initial threshold.

In this embodiment, the first threshold setting unit is configured to increase a preset offset based on the initial threshold to obtain the first threshold. In some other embodiments, it may also be: the first threshold setting unit is configured to use the initial threshold as the first threshold. In other embodiments, it may also be that: the first threshold setting unit is configured to multiply the second threshold by a second scaling factor to obtain the first threshold.

It should be noted that any pixel point 110 has a corresponding grayscale gradient in the X direction and Y direction, therefore, as an example, the relational expression of the grayscale gradient of the pixel point includes:

Wherein, M(x, y) represents the gray gradient of the pixel point 110, gx represents the gradient of the pixel point 110 in the X direction, gy represents the gradient of the pixel point 110 in the Y direction, and the X The direction is the row direction of the pixel array, and the Y direction is the column direction of the pixel array.

It should also be noted that the preset offset corresponding to the first threshold should not be too small, nor should it be too large. If the preset offset is too small, it is easy to cause the false detection rate to become high when performing the first identification process, that is, it is easy to classify the normal pixel point 110 as the initial abnormal point 200, resulting in the initial abnormal point 200 If the number is too large, the amount of subsequent data processing for obtaining abnormal points according to the initial abnormal points and the second identification process will be increased; if the preset offset is too large, the probability of missed detection will easily increase. Therefore, in this embodiment, the preset offset is 3-5.

In addition, only one pixel point 110 is shown as the initial abnormal point 200 in FIG. 5 , but the initial abnormal point 200 is not limited to one pixel point. For example, in other embodiments, after the first identification process, multiple initial abnormal points are obtained, and there is a situation that multiple initial abnormal points are connected.

The abnormal point obtaining module 70 is used for obtaining the abnormal point 250 according to the initial abnormal point 200 .

Subsequently, by further identifying the abnormal point 250, it is judged whether the abnormal point 250 is a defect point or a noise point.

In this embodiment, the abnormal point acquisition module 70 includes: a connected domain judgment unit, configured to perform connected domain judgment on the initial abnormal point 200 in the image to be processed 100, and obtain a connected component (Connected Component), the same connected domain Each initial outlier point 200 in has adjacent initial outlier points 200, and the initial outlier points 200 in the connected domain and outside the connected domain are separated from each other; the third screening subunit is used for the initial outlier points 200 in the connected domain When the number is greater than or equal to the preset number, determine the initial abnormal point 200 in the connected domain as a defect point; domain as the outlier 250.

A collection of multiple connected pixel points 110 meeting certain conditions is called a connected domain. Specifically in this embodiment, the specific condition is to meet the first threshold condition.

The outlier point 250 is rechecked in the second identification process. Therefore, it is determined whether the second identification process needs to be performed by first performing connected domain judgment.

Specifically, in the case where it is difficult to determine whether the type of the initial outlier 200 is a defect point or an outlier 250, the connected domain is judged first, so that the number of initial outlier points 200 in the connected domain is greater than or equal to the preset number In the case of , it can be determined that the initial abnormal point 200 is a defect point, and then there is no need to select surrounding pixels and perform the second identification process, so that the number of initial abnormal points 200 in the connected domain can be selectively less than In the case of a preset number, the abnormal points 250 are re-inspected, thereby improving the detection efficiency while improving the accuracy of the detection results.

In this embodiment, the determination of connected domains includes determination of four connected domains or determination of eight connected domains. Among them, the four-connected domain judgment refers to: judging whether there are 4 adjacent initial abnormal points 200 around any initial abnormal point 200; the eight-connected domain judgment refers to: judging whether there are 8 adjacent initial abnormal points 200 Adjacent initial outliers 200.

When the number of 200 initial abnormal points in the connected domain is greater than or equal to the preset number, the initial abnormal points 200 in the connected domain are used as defect points, and the number of 200 initial abnormal points in the connected domain is less than the preset number In the case of a large number, the connected domain is used as the abnormal point 250, and the pixel size of the noise point is usually small. Therefore, if the preset number is too large, it is easy to misjudge the defective point as the abnormal point when judging the connected domain. Abnormal point 250, so unnecessary follow-up operations are performed. Therefore, in this embodiment, the preset number is 1 to 5.

As shown in FIG. 5 , as an example, the preset number is one. Correspondingly, if the initial outlier point 200 is an isolated pixel point 110, the initial outlier point 200 is taken as the outlier point 250; otherwise, the initial outlier point 200 is taken as a defect point. The pixel size of the noise point is generally small, usually one pixel point, therefore, by setting the preset number to one, it is beneficial to accurately obtain the abnormal point 250 and reduce the computation load of subsequent operations.

It should be noted that, when the number of initial outliers in the connected domain is less than the preset number, determining the connected domain as the outlier refers to taking all the initial outliers in the connected domain as one Outlier.

In other embodiments, the abnormal point acquisition module includes: an adjacent point acquisition subunit, configured to acquire pixel points adjacent to the initial abnormal point as adjacent points, and the number of the adjacent points is one or more ; The fourth screening subunit is used to judge whether each adjacent point is an initial abnormal point, and if none of the adjacent points is an initial abnormal point, take the initial abnormal point as an abnormal point.

Specifically, the number of the adjacent points may be four or eight. When the number of the adjacent points is four, the initial abnormal point and the adjacent points are in a cross-shaped layout or an X-shaped layout; when the number of the adjacent points is eight, the initial abnormal point The abnormal points and the adjacent points are arranged in a 3*3 array.

The pixel point selection module 30 is used to select one or more pixel points 110 adjacent to the abnormal point 250 as surrounding pixel points 300 .

The second identification module 40 performs a second identification process on the surrounding pixel points 300 , so as to realize the re-examination of the abnormal points 200 .

The pixel size of noise is usually small (for example, the pixel size of one noise is one pixel), which makes it difficult to distinguish defects from noise. However, the pixels adjacent to the noise point are all normal pixels, and the pixels adjacent to the real defect are usually also defective pixels. Therefore, in the detection process, when abnormal points 250 appear, the second threshold condition is used to The surrounding pixel points 300 are identified again, the second threshold condition includes the first threshold condition, that is, the set corresponding to the first threshold condition is included in the set corresponding to the second threshold condition, which makes In the case where the difference between the surrounding pixel 300 and the normal pixel is small, it can still be detected, so that more defective pixels can be detected around the abnormal point 250, and correspondingly, noise points and defective points can be better distinguished. By accurately judging whether the abnormal point 250 is a real defect point, it is convenient to eliminate noise points and retain the defect points, correspondingly reducing the probability of false detection and missed detection, thereby improving the accuracy of detection results.

In this embodiment, by selecting the pixel point 110 close to the abnormal point 250 as the surrounding pixel point 300, the correlation between the detection result of the surrounding pixel point 300 and the abnormal point 250 is higher, thereby improving the subsequent 2. The accuracy of the detection results of the identification process.

In this embodiment, the pixel point selection module 30 is used to select the pixel points 110 adjacent to the edge and/or vertex corner of the abnormal point 250 as surrounding pixel points 300 . Specifically, the number of the surrounding pixel points 300 is 8, and the surrounding pixel points 300 and the abnormal point 250 form a 3*3 pixel array. That is to say, the abnormal point 250 is the central pixel point in the 3*3 pixel array, and the surrounding pixel point 300 is the pixel point 110 next to the abnormal point 250 . The abnormal point 250 is the central pixel point in the 3*3 pixel array, therefore, by selecting 8 surrounding pixel points 300, it is beneficial to improve the result accuracy of the second recognition process.

In other embodiments, the number of surrounding pixel points is four, and the surrounding pixel points and the abnormal point are in a cross-shaped layout or an X-shaped layout.

It should be noted that, in the third screening subunit of the abnormal point acquisition module 70, when the number of initial abnormal points in the connected domain is less than 200, the connected domain is determined as the abnormal point. 250 points. Wherein, when the preset number is multiple, determining the connected domain as the outlier point 250 refers to: taking all initial outlier points 200 in the connected domain as one outlier point 250 . Correspondingly, the pixel point selection module 30 is used to select the pixel points 110 adjacent to the edge and/or vertex corner of the abnormal point 250 as the surrounding pixel points 300, and the surrounding pixel points of the abnormal point 250 are 300 The number may be greater than 8, for example, when the outlier 250 includes two initial outliers 200, the number of surrounding pixels 300 is 10 or 6.

It should also be noted that, in other embodiments, when the abnormal point is obtained by obtaining pixels adjacent to the initial abnormal point as adjacent points, the pixel point selection module is used to use the initial abnormal point Adjacent points are used as the surrounding pixel points of the initial abnormal point.

The second identification module 40 is configured to perform a second identification process on the surrounding pixels 300, and use the surrounding pixels 300 satisfying a second threshold condition as defective pixels, the second threshold condition including the first threshold condition; The judging module 50 is configured to determine that the abnormal point 200 is a defective point when there is a defective pixel point in the surrounding pixel point 300, and determine that the abnormal point 200 is a defective point when there is no defective pixel point in the surrounding pixel point 300. The abnormal points 200 are noise points.

Here, the second threshold condition refers to: a set of surrounding pixel points 300 meeting the preset condition.

The second threshold condition is used as a criterion for judging whether the surrounding pixel point 300 is a defective pixel point, that is, if the surrounding pixel point 300 meets the second threshold value condition, the surrounding pixel point 300 is regarded as a defective pixel point .

In this embodiment, the second threshold condition includes the first threshold condition, that is, the set corresponding to the first threshold condition is included in the set corresponding to the second threshold condition, so that the second identification process detects The number of pixels that are different from normal pixels is increased.

If the abnormal point 200 is a defect point, there is usually a difference between the surrounding pixel points 300 around the abnormal point 200 and the normal pixel points 110, but the difference is small, and the pixels around the noise point are all Normal pixels, therefore, use the second threshold condition to carry out the second identification process, if the abnormal point 200 is a defect point, as shown in Figure 7 (a), after the second identification process, it can also be in the One or more defective pixel points are detected around the abnormal point 200. For example, FIG. If the abnormal point 200 is a noise point, as shown in FIG. 7(b), after the second identification process, no defective pixel points can be detected around the abnormal point 200, so that the true value of the abnormal point 200 Types are distinguished, that is, noise points and defect points are distinguished, so that the noise points can be removed later and the defect points can be retained.

In this embodiment, the second identification module 40 includes: a second threshold setting unit for setting a second threshold; a third image selection unit for acquiring a reference image; a second matching unit for matching the reference The image is matched with the image to be processed 100, so that the pixels in the matching area of the image to be processed and the reference image are in one-to-one correspondence; the third comparison unit is used to compare the surrounding pixels 300 with the reference Comparing corresponding pixel points in the image to obtain a second difference value between the intensity representation values of the surrounding pixel point 300 and the corresponding pixel point in the reference image 152; the fourth comparison unit is used to compare the first Two difference values and a second threshold value, acquiring peripheral pixel points 300 whose second difference values are greater than the second threshold value as second difference points; a second abnormal point acquisition unit, configured to use the second difference points to obtain defective pixel points .

Correspondingly, in this embodiment, the second threshold condition at least includes: the second difference value is greater than a second threshold, wherein the second threshold is smaller than the first threshold.

In this embodiment, by making the second threshold smaller than the first threshold, the second threshold condition includes the first threshold condition.

Specifically, the second abnormal point acquisition unit includes: a second difference number acquisition subunit, configured to acquire the number of the second difference points to obtain a second difference number, and a second screening subunit, used to obtain the second difference number in the In the case where the second difference quantity is greater than or equal to the preset quantity, it is determined that the currently detected surrounding pixel point 300 is a defective pixel point.

Correspondingly, the second threshold condition further includes: the second difference quantity is greater than or equal to a preset quantity.

Here, the minimum value of the preset number is one, and the maximum value is the total number of reference images 152 corresponding to the current surrounding pixel point 300 to be measured. The preset number can be set according to actual needs.

It should be noted that, in other embodiments, the second abnormal point obtaining unit may also directly use the second difference point as a defective pixel point.

It should also be noted that the image to be tested 151 recognized by the second recognition module 40 is the unit image 150 where the abnormal point 200 is located.

In this embodiment, the second threshold is positively correlated with the definition of the image 100 to be processed.

In this embodiment, the first threshold setting unit is configured to increase a preset offset based on the initial threshold to obtain the first threshold, and the second threshold setting unit is configured to increase a preset offset based on the initial threshold The offset is to obtain the second threshold, and the preset offset corresponding to the first threshold is different from the preset offset corresponding to the second threshold.

Specifically, the second threshold is smaller than the first threshold, so that the second threshold condition includes the first threshold condition.

In other embodiments, it may also be: when the first threshold setting unit is used to set the initial threshold as the first threshold, the second threshold setting unit is used to set the first threshold and The first scaling factor is multiplied to obtain the second threshold. In other embodiments, it may also be: when the first threshold setting unit is used to multiply the second threshold by a second scaling factor to obtain the first threshold, the second threshold setting unit Used to use the initial threshold as the second threshold.

It should be noted that, when the second threshold is smaller than the first threshold, the ratio of the second threshold to the first threshold should not be too small, nor should it be too large. If the ratio of the second threshold to the first threshold is too small, it is easy to cause the second threshold to be too small, and when the second identification process is performed, it is easy to cause the false detection rate to become high, that is to say, it is easy to Classify the normal surrounding pixels 300 as defective pixels, so that it is easy to mistakenly classify the abnormal points 200 as defective points; if the ratio of the second threshold to the first threshold is too large, it is easy to increase the probability of missed detection , so it is easy to misclassify the abnormal point 200 as a noise point. Therefore, in this embodiment, the second threshold is 60% to 80% of the first threshold.

The judgment module 50 makes a judgment based on the detection result of the second recognition module 40 . Specifically, if there is a defective pixel in the surrounding pixels 300, it is determined that the abnormal point 200 is a defective point; when there is no defective pixel in the surrounding pixels 300, it is determined that the abnormal Point 200 is noise.

It should be noted that, in other embodiments, in the first identification module, the third outlier acquisition unit is used instead of the first comparison unit, the second comparison unit and the first outlier acquisition unit; In the identification module, the fourth abnormal point acquisition unit is used to replace the third comparison unit, the fourth comparison unit and the second abnormal point acquisition unit.

For example, in the case that the image to be processed is a dark field image, in the first identification module, the third outlier acquisition unit is configured to compare the pixels of the image to be processed with the first threshold In comparison, the pixel point whose intensity characteristic value is greater than the first threshold is acquired as an initial abnormal point; the first threshold condition includes: the intensity characteristic value of the pixel point is greater than the first threshold.

Correspondingly, in the second identification module, the fourth outlier acquisition unit is configured to compare the surrounding pixel points with a second threshold, and acquire the outlier whose intensity characteristic value is greater than the second threshold. The surrounding pixels are regarded as defective pixels; the second threshold condition includes: the intensity characteristic value of the surrounding pixels is greater than the second threshold; wherein the second threshold is smaller than the first threshold.

Alternatively, when the image to be processed is a bright field image, in the first identification module, the third outlier acquisition unit is configured to compare each pixel of the image to be processed with the first threshold For comparison, the pixel points whose intensity characteristic value is smaller than the first threshold are acquired as initial abnormal points; the first threshold condition includes: the intensity characteristic value of the pixel point is smaller than the first threshold.

Correspondingly, in the second identification module, the fourth outlier acquisition unit is configured to compare the surrounding pixel points with a second threshold, and acquire the outlier whose intensity characteristic value is smaller than the second threshold. The surrounding pixels are regarded as defective pixels; the second threshold condition includes: the intensity characteristic value of the surrounding pixels is smaller than the second threshold; wherein the second threshold is greater than the first threshold.

During the first identification process, acquire the pixel points whose intensity characteristic value is greater than the first threshold value as the initial abnormal point; during the second identification process, acquire the surrounding area whose intensity characteristic value is greater than the second threshold value A pixel is a defective pixel, therefore, by making the second threshold smaller than the first threshold, the second threshold condition includes the first threshold condition.

It should be noted that if only the first identification module 20 is set in the detection system, and the first identification module 20 directly uses the second threshold condition to perform the first identification process, it will easily lead to a high false detection rate (for example, there will be A certain intensity characteristic value difference, but the pixel point whose intensity characteristic value difference is within an acceptable range is classified as a defect point), therefore, in this embodiment, an abnormal point acquisition module 70 and a pixel point selection module 30 are also set in the detection system , the second identification module 40 and the judgment module 50, first use the first threshold condition to perform the first identification process, obtain the initial abnormal point 200, and obtain the abnormal point 250 according to the initial abnormal point 200, to exclude normal pixel points 110 or The pixel points 110 whose intensity characteristic value difference is within the acceptable range or the pixel points 110 determined to be defective points, so that the number of abnormal points 250 will not be too many, and then use the second threshold value condition to perform the second identification process on the surrounding pixel points 300 , so as to realize the re-inspection of the abnormal point 250, so that the real defect point can be screened out more accurately while reducing the data processing amount of the second identification process.

An embodiment of the present invention also provides a device, which can realize the detection method provided by the embodiment of the present invention by loading the above detection method in the form of a program.

Referring to FIG. 7 , it shows a hardware structural diagram of a device provided by an embodiment of the present invention. The device described in this embodiment includes: at least one processor 01 , at least one communication interface 02 , at least one memory 03 and at least one communication bus 04 .

In this embodiment, the number of the processor 01, the communication interface 02, the memory 03, and the communication bus 04 is at least one, and the processor 01, the communication interface 02, and the memory 03 complete the mutual communication through the communication bus 04. Communication.

The communication interface 02 may be an interface of a communication module for network communication, for example, an interface of a GSM module.

The processor 01 may be a central processing unit CPU, or an ASIC (Application Specific Integrated Circuit), or one or more integrated circuits configured to implement the detection method described in this embodiment.

The memory 03 may include a high-speed RAM memory, and may also include a non-volatile memory (non-volatile memory), such as at least one disk memory.

Wherein, the memory 03 stores one or more computer instructions, and the one or more computer instructions are executed by the processor 01 to implement the detection method provided by the foregoing embodiments.

It should be noted that the above-mentioned implementation terminal device may also include other devices (not shown) that may not be necessary for the disclosure of the embodiments of the present invention; in view of the fact that these other devices may not be necessary for understanding the disclosure of the embodiments of the present invention, This embodiment of the present invention does not introduce them one by one.

An embodiment of the present invention also provides a storage medium, the storage medium stores one or more computer instructions, and the one or more computer instructions are used to implement the detection method provided by the foregoing embodiments.

In the detection method of this embodiment, the first recognition process is performed on the image to be processed, and after obtaining the pixel points satisfying the first threshold condition in the image to be processed as the initial abnormal point, the abnormal point is obtained according to the initial abnormal point, and an abnormal point is selected. or a plurality of pixels adjacent to the abnormal point are regarded as surrounding pixels, and the surrounding pixels are subjected to the second identification process, and the surrounding pixels satisfying the second threshold condition are regarded as defective pixels, and the second The threshold condition includes the first threshold condition; wherein, the pixels adjacent to the noise point are normal pixels, and the pixels adjacent to the real defect are usually defective pixels, therefore, in the detection process, when a After the abnormal point, use the second threshold condition to identify the surrounding pixel points again. The second threshold condition includes the first threshold condition, that is, the set corresponding to the first threshold condition is included in the second threshold condition. In the set corresponding to the threshold condition, this makes it possible to detect when the difference between the surrounding pixels and the normal pixels is small, so that more defective pixels can be detected around the abnormal points, and the corresponding can be better. Distinguish between noise points and defect points to accurately judge whether the abnormal points are real defect points, so as to remove noise points and retain defect points, correspondingly reduce the probability of false detection and missed detection, and improve the accuracy of detection results.

The embodiments of the present invention described above are combinations of elements and features of the present invention. The elements or features may be considered optional unless mentioned otherwise. Each element or feature may be practiced without being combined with other elements or features. In addition, the embodiments of the present invention may be configured by combining some elements and/or features. The order of operations described in the embodiments of the present invention may be rearranged. Some constructions of any one embodiment may be included in another embodiment, and may be replaced with corresponding constructions of another embodiment. It is obvious to those skilled in the art that claims that have no explicit citation relationship with each other among the appended claims may be combined in the embodiments of the present invention or may be included as new claims in amendments after filing the present application.

Embodiments of the present invention can be realized by various means such as hardware, firmware, software, or a combination thereof. In the hardware configuration mode, the method according to the exemplary embodiment of the present invention can be implemented by one or more application-specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices ( PLD), field programmable gate array (FPGA), processor, controller, microcontroller, microprocessor, etc. to achieve.

In a firmware or software configuration, the embodiments of the present invention can be implemented in the form of modules, procedures, functions, and the like. The software codes may be stored in memory units and executed by processors. The memory unit is located inside or outside the processor, and can transmit data to and receive data from the processor via various known means.

The above description of the disclosed embodiments is provided to enable any person skilled in the art to make or use the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the present invention will not be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Although the present invention is disclosed above, the present invention is not limited thereto. Any person skilled in the art can make various changes and modifications without departing from the spirit and scope of the present invention, so the protection scope of the present invention should be based on the scope defined in the claims.

Claims (17)

1. A detection method, characterized in that, comprising: Acquiring an image to be processed, the image to be processed has a plurality of pixels; performing a first recognition process on the image to be processed, and acquiring pixels satisfying a first threshold condition in the image to be processed as initial abnormal points; Obtaining abnormal points according to the initial abnormal points; Selecting one or more pixel points adjacent to the abnormal point as surrounding pixel points; Performing a second identification process on the surrounding pixels, using surrounding pixels that meet a second threshold condition as defective pixels, the second threshold condition including the first threshold condition; If there is a defective pixel in the surrounding pixels, it is determined that the abnormal point is a defective point, and in the case that there is no defective pixel in the surrounding pixels, it is determined that the abnormal point is a noise point. 2. The detection method according to claim 1, wherein said performing the first identification process on the image to be processed comprises: acquiring a reference image; performing matching processing on the reference image and the image to be processed, Making the pixel points in the matching area of the image to be processed and the image to be processed correspond one-to-one; comparing the reference image with the image to be processed, and obtaining corresponding pixels in the reference image and the image to be processed The first difference value between the intensity characteristic values; compare the first difference value with the first threshold value, obtain the pixel point of the image to be processed whose first difference value is greater than the first threshold value as the first difference point; use the The first difference point obtains an initial abnormal point; the first threshold condition at least includes: the first difference value is greater than a first threshold; The second identification process on the surrounding pixels includes: acquiring a reference image; performing matching processing on the reference image and the image to be processed, so that the image to be processed and the reference image match the pixel points of the region One-to-one correspondence; comparing the surrounding pixels with the corresponding pixels in the reference image, and acquiring a second difference value between the intensity representation values of the surrounding pixels and the corresponding pixels in the reference image ; Comparing the second difference value with a second threshold value, acquiring the surrounding pixel points whose second difference value is greater than the second threshold value as the second difference point; using the second difference point to obtain a defective pixel point; the second difference point The threshold condition at least includes: the second difference value is greater than a second threshold, wherein the second threshold is smaller than the first threshold; The intensity characteristic value is positively correlated with the gray value or the signal-to-noise ratio. 3. The detection method according to claim 1, wherein the image to be processed is a dark field image; The first identification process on the image to be processed includes: comparing the pixels of the image to be processed with a first threshold, and acquiring the pixel whose intensity characteristic value is greater than the first threshold as an initial abnormality point; the first threshold condition includes: the intensity characteristic value of the pixel point is greater than the first threshold; The second identification process on the surrounding pixels includes: comparing the surrounding pixels with a second threshold, and obtaining the surrounding pixels whose intensity characteristic value is greater than the second threshold as defective pixels; The second threshold condition includes: the intensity characteristic value of the surrounding pixel points is greater than the second threshold; wherein the second threshold is smaller than the first threshold; The intensity characteristic value is positively correlated with the gray value or the signal-to-noise ratio. 4. The detection method according to claim 1, wherein the image to be processed is a bright field image; The first identification process on the image to be processed includes: comparing the pixels of the image to be processed with a first threshold, and acquiring the pixel whose intensity characteristic value is smaller than the first threshold as an initial abnormality point; the first threshold condition includes: the intensity characteristic value of the pixel point is less than the first threshold; The second identification process on the surrounding pixels includes: comparing the surrounding pixels with a second threshold, and obtaining the surrounding pixels whose intensity characteristic value is smaller than the second threshold as defective pixels; The second threshold condition includes: the intensity characteristic value of the surrounding pixel points is less than the second threshold; wherein the second threshold is greater than the first threshold; The intensity characteristic value is positively correlated with the gray value or the signal-to-noise ratio. 5. The detection method according to claim 2, wherein the image to be processed comprises a plurality of identical unit images; performing the first recognition process on each of the unit images, acquiring the initial abnormal points of each unit image, using the unit image currently to be tested as the image to be tested, and taking a plurality of the units around the image to be tested image as a reference image; The acquiring initial abnormal points by using the first difference points includes: acquiring the number of the first difference points to obtain a first difference number, and when the first difference number is greater than or equal to a preset number, determine The currently detected pixel point is an initial abnormal point; the first threshold condition further includes: the first difference number is greater than or equal to a preset number; The acquiring defective pixel points by using the second difference points includes: acquiring the number of the second difference points to obtain a second difference number, and when the second difference number is greater than or equal to a preset number, determine The currently detected surrounding pixels are defective pixels; the second threshold condition further includes: the second difference amount is greater than or equal to a preset number. 6. The detection method according to any one of claims 2-4, wherein the first threshold and the second threshold are positively correlated with the definition of the image to be processed. 7. The detection method according to claim 6, wherein acquiring the first threshold and/or acquiring the second threshold comprises an initial threshold acquisition process, and the initial threshold acquisition process comprises: acquiring the pending threshold The grayscale gradient of each pixel in the image; obtaining the average value of the grayscale gradient of the image to be processed, and using the average value of the grayscale gradient as an initial threshold; The detection method further includes: adding a preset offset based on the initial threshold. 8. The detection method according to claim 7, wherein obtaining the first threshold further comprises: using the initial threshold as the first threshold; Obtaining the second threshold includes: multiplying the first threshold by a first scaling factor to obtain the second threshold; or, Acquiring the second threshold further includes: using the initial threshold as the second threshold; Obtaining the first threshold includes: multiplying the second threshold by a second scaling factor to obtain the first threshold; or, Both acquiring the first threshold and acquiring the second threshold include initial threshold acquisition processing, and the preset offset corresponding to the first threshold is different from the preset offset corresponding to the second threshold. 9. The detection method according to claim 7, wherein the first threshold is obtained by increasing a preset offset based on the initial threshold; the second threshold is smaller than the first threshold; The preset offset corresponding to the first threshold is 3 to 5; the second threshold is 60% to 80% of the first threshold. 10. detection method as claimed in claim 7 is characterized in that, the relational expression of the gray scale gradient of described pixel comprises:

Among them, M(x, y) represents the gray gradient of the pixel, gx represents the gradient of the pixel in the X direction, gy represents the gradient of the pixel in the Y direction, and the X direction is the pixel The row direction of the array, and the Y direction is the column direction of the pixel array. 11. The detection method according to claim 1, wherein obtaining the abnormal points according to the initial abnormal points comprises: judging the connected domains of the initial abnormal points in the image to be processed, obtaining the connected domains, and the same connected domains Each initial outlier point in has adjacent initial outlier points, and the initial outlier points in the connected domain and outside the connected domain are separated from each other; When the number of initial abnormal points in the connected domain is greater than or equal to the preset number, the initial abnormal points in the connected domain are used as defect points, and in the case where the number of initial abnormal points in the connected domain is less than the preset number Next, the connected domain is used as the outlier point; or, Obtaining an abnormal point according to the initial abnormal point includes: acquiring pixel points adjacent to the initial abnormal point as adjacent points, and the number of the adjacent points is one or more; It is judged whether each adjacent point is an initial outlier point, and if none of the adjacent points is an initial outlier point, the initial outlier point is taken as an outlier point. 12. The detection method according to claim 11, wherein the preset number is 1 to 5. 13. The detection method according to claim 1, wherein the number of the surrounding pixel points is 8, and the surrounding pixel points and the abnormal point form a 3*3 pixel array; Alternatively, the number of surrounding pixels is four, and the surrounding pixels and the abnormal point are in a cross-shaped layout or an X-shaped layout. 14. The detection method according to claim 1, wherein the acquiring the image to be processed comprises: providing an imaging system and the object to be tested; using the imaging system to capture an image of the object to be tested as the Process images. 15. A detection system, characterized in that it comprises: An image acquisition module, configured to acquire an image to be processed, the image to be processed has a plurality of pixels; The first recognition module is configured to perform a first recognition process on the image to be processed, and acquire pixels satisfying a first threshold condition in the image to be processed as initial abnormal points; An abnormal point obtaining module, configured to obtain an abnormal point according to the initial abnormal point; A pixel selection module, configured to select one or more pixels adjacent to the abnormal point as surrounding pixels; The second identification module is configured to perform a second identification process on the peripheral pixels, and use peripheral pixels satisfying a second threshold condition as defective pixels, where the second threshold condition includes the first threshold condition; A judging module, configured to determine that the abnormal point is a defective point when there is a defective pixel point among the surrounding pixels, and determine the abnormal point when there is no defective pixel point among the surrounding pixel points for noise. 16. A device comprising at least one memory and at least one processor, the memory storing one or more computer instructions, wherein the one or more computer instructions are executed by the processor to implement The detection method according to any one of claims 1 to 14. 17. A storage medium, characterized in that the storage medium stores one or more computer instructions, and the one or more computer instructions are used to implement the detection method according to any one of claims 1 to 14.

Images