CN114755243A - A method for detecting surface cracks of FPC connectors based on machine vision - Google Patents

Abstract

The invention discloses a method for detecting surface cracks of FPC connectors based on machine vision, which relates to the technical field of semiconductors. The element performs multi-scale morphological operations on the image to be processed to obtain a two-dimensional crack probability map, and uses the segmentation threshold to perform global adaptive threshold segmentation. The candidate crack region at the position of the target connected domain is located in the dimensional crack probability map; according to the size and position information of the target connected domain and the response intensity value of the corresponding candidate crack region, it can be detected whether there is a crack on the surface of the FPC connector to be detected. The method realizes the automatic detection of surface cracks through machine vision, and has many advantages such as high efficiency, non-contact, high precision, fast response speed, and adaptability to non-optical harsh environments.

Description

A method for detecting surface cracks of FPC connectors based on machine vision

technical field

The invention relates to the technical field of semiconductors, in particular to a method for detecting surface cracks of FPC connectors based on machine vision.

Background technique

The surface mount connector is a structure that is connected to the printed circuit board (PCB) in a mountable installation method. It can meet the process requirements of surface mount technology (SMT) automatic sticking and reflow soldering, and can greatly improve the overall performance of the machine. Assembly efficiency, increase the installation density, save the space of the whole machine.

With the popularity of SMT technology, the application of surface mount connectors has become more and more extensive, and various types of PCBs have corresponding surface mount connectors. FPC connector is a connector used to connect flexible circuit board (FPC) and ordinary PCB. FPC is also called flexible circuit board and flexible circuit board. It has excellent characteristics such as light weight, thin thickness, and free bending and folding. and favored. FPC connectors are often used for the connection of liquid crystal display (LCD) to the driving circuit PCB, mainly 0.5mm pitch products, and 0.3mm pitch products have also been widely used.

With the expansion of the production scale of FPC connectors, the requirements for product quality are also rapidly increasing, which leads to product quality inspection becoming a crucial link in the connector production steps. With the emergence of new technologies and new materials, connectors have begun to miniaturize and thin the appearance size, and the high-precision trend of wire grooves and main bodies has developed, which has put forward higher requirements for the quality inspection of connectors. To meet the requirements of the SMT process, the plastic end face of the entire product is required to have good flatness and coplanarity. If the plastic end face has cracks, it will lead to poor connection between the connector and the cable, thus affecting the use of the product.

At present, most factories still use the traditional manual inspection method, and experienced employees classify the batches of products to be inspected through their knowledge of the products. The manual detection method is simple and the investment is small, but it mainly relies on manual visual recognition and manual operation, which is inefficient, has a high possibility of misjudgment and lacks a certain degree of objectivity.

SUMMARY OF THE INVENTION

In view of the above problems and technical requirements, the present inventor proposes a method for detecting surface cracks of FPC connectors based on machine vision. The technical solution of the present invention is as follows:

A method for detecting surface cracks of FPC connectors based on machine vision, the method includes:

Obtain the to-be-processed image F(x,y) of the to-be-detected area of the to-be-detected FPC connector;

The two-dimensional crack probability map P(x,y) is obtained by performing multi-scale morphological operations on the to-be-processed image F(x,y) using the structural elements S(x,y) of various typical surface crack characteristics;

Use the segmentation threshold t to perform global adaptive threshold segmentation on the two-dimensional crack probability map P(x,y), and obtain the crack binary map Q(x,y);

Search and mark the connected domain of the crack binary map Q(x,y), determine the target connected domain that satisfies the preset connected domain characteristics, and determine the position of the target connected domain in the two-dimensional crack probability map P(x,y) The response intensity value mean of the candidate crack region at ;

If the size and position information of the target connected domain and the response intensity value mean of the corresponding candidate crack region both satisfy the corresponding preset conditions, it is determined that there is a crack on the surface of the FPC connector to be detected.

Its further technical solution is to obtain the to-be-processed image of the to-be-detected area of the to-be-detected FPC connector, including:

Obtain the original image of the region of interest of the FPC connector to be detected. The region of interest includes the area between the upper and lower boundaries of the FPC connector to be detected in the vertical column direction, and the left boundary of the region of interest in the horizontal row direction is the region to be detected. Detect the first metal pin of the FPC connector, and the right border is the last metal pin of the FPC connector to be detected;

The original image of the region of interest is processed into a binary image and the coordinates of the lower boundary pixel are determined. The lower boundary pixel is located at the lower boundary of the FPC connector to be detected in the binary image and is adjacent to the pixel point. Pixels with different pixel values;

Perform straight line fitting based on the coordinates of the lower boundary pixel points in the binary image to obtain the lower boundary fitting line, and translate the lower boundary fitting line upward according to the width d of the area to be detected to obtain the upper boundary fitting line;

The image to be processed is obtained by intercepting the image between the upper boundary fitting line and the lower boundary fitting line from the original image of the region of interest.

Its further technical solution is to perform straight line fitting based on the coordinates of the lower boundary pixel points in the binary image to obtain the lower boundary fitting line, including:

The coordinates of any two lower boundary pixel points are fitted with a straight line to obtain a candidate fitting line, and the number of lower boundary pixel points whose distance from the candidate fitting line is within the error range is determined as the fitting line of the candidate fitting line. the exact number of data points;

Re-execute the steps of performing straight line fitting on the coordinates of any two lower boundary pixel points to obtain a candidate fitting line until the iteration termination condition is reached, and select the candidate fitting line obtained by each iteration with the largest number of corresponding fitting accurate data points. A candidate fit line serves as the lower bound fit line.

Its further technical solution is to process the original image of the region of interest into a binary image and determine the coordinates of the location of the lower boundary pixel point, including:

Convert the original image of the region of interest of the RGB three-channel into the image of the HSV three-pass, and extract the V channel as the preprocessed grayscale image;

Threshold segmentation is performed on the preprocessed grayscale image to obtain a binary image, and the coordinates of the position of the lower boundary pixel point where the pixel value of the adjacent pixel point in the lower half of the binary image is abruptly changed are extracted.

Its further technical solution is to use the structural elements S(x,y) of various typical surface crack characteristics to perform multi-scale morphological operations on the to-be-processed image F(x,y) to obtain a two-dimensional crack probability map P(x,y) , including obtaining a two-dimensional crack probability map P(x,y) as:

P(x,y)=max{(F·S _i )(x,y)}-F(x,y);

Among them, S _i (x,y) is the structuring element of the i-th typical surface crack feature, (F·S _i )(x,y)-F(x,y) represents the gray-level black hat operation, (F· S _i )(x,y) represents the gray level closing operation and

Its further technical scheme is, the method also includes:

The two-dimensional crack probability map P(x,y) is projected and transformed in the longitudinal column direction to obtain a one-dimensional crack probability map P'(x,y);

Perform Gaussian filtering on the one-dimensional crack probability map P'(x, y) according to the preset interval h, and determine the central pixel column corresponding to the maximum value in the filtering result;

Extract the local area f(x,y) in the two-dimensional crack probability map P(x,y) with the center pixel column including several pixel columns on both sides in the center;

The segmentation threshold t is obtained by solving the maximum inter-class variance method in the local area f(x, y).

Its further technical scheme is, when the fitting path length l of a connected domain is greater than the length threshold, and the area reaches the area threshold, and the included angle between the main direction and the horizontal row direction reaches the angle threshold, it is determined that the connected domain satisfies the preset connectivity. The target connected domain of the domain feature.

Its further technical solution is to determine the response intensity value mean of the candidate crack region at the position of the target connected domain in the two-dimensional crack probability map P(x,y), including:

Determine the candidate crack region at the target connected domain in the two-dimensional crack probability map P(x, y), and use the gray average value of the pixels contained in the candidate crack region as the response intensity value mean of the candidate crack region.

Its further technical scheme is that the size and position information of the target connected domain includes the fitting path length l of the target connected domain, the first distance _Lu between the upper end point of the target connected domain and the upper boundary of the area to be detected, and the target connected domain. The second distance L _d between the lower end point of and the lower boundary of the area to be detected.

Its further technical scheme is, when l>L and at least one of L _u and L _d reaches the corresponding distance threshold, determine that the size and position information of the target connected domain satisfies the corresponding preset condition; when mean>M, determine the candidate. The response intensity value mean of the crack area satisfies the corresponding preset conditions, L is the fitting path length threshold, and M is the response intensity threshold.

The beneficial technical effects of the present invention are:

The present application discloses a method for detecting surface cracks of FPC connectors based on machine vision. The method realizes automatic detection of surface cracks through machine vision, and has many advantages such as non-contact, high precision, fast response speed, and adaptability to non-optical harsh environments. It can work under relatively harsh conditions, and will not cause damage to the product during inspection, nor will it introduce secondary defects in the inspection process, and has faster speed and greater stability than manual inspection. and repeatability, while also improving enterprise productivity and reducing labor costs.

The method uses the method of dividing the region of interest to roughly locate the area to be detected after obtaining the original image of the entire area, and then accurately extracts the lower boundary of the area to be detected by the method to complete the accurate fitting of the upper and lower boundaries. The upper and lower boundaries are precisely located in the area of interest to the actual area to be detected, with high accuracy.

In the area to be inspected, the crack probability map of the area to be inspected is obtained by the proposed multi-scale morphological operation method, and the threshold segmentation of the above probability map is completed based on the proposed adaptive threshold segmentation method. Connected domain analysis and connected domain screening, retain the suspected crack area, and finally complete the detection of cracks in the area to be detected by setting the threshold of crack-related characteristics, which can meet the real-time online detection of surface cracks of various types of FPC connectors, and has high detection accuracy.

Description of drawings

FIG. 1 is a schematic flowchart of a method for detecting surface cracks of an FPC connector according to an embodiment.

FIG. 2 is a schematic diagram of a to-be-processed image F(x, y) of a to-be-detected area obtained by step-by-step processing of a full-area original image in an embodiment.

Figure 3 shows the structural elements of six typical surface crack features.

FIG. 4 is a schematic flowchart of a method for detecting surface cracks of an FPC connector according to another embodiment.

FIG. 5 is a schematic diagram of a two-dimensional crack probability map P(x,y) in one embodiment.

FIG. 6 is a schematic diagram of the crack binary map Q(x, y) obtained by performing global adaptive threshold segmentation on FIG. 5 .

FIG. 7 is a schematic diagram of cracks on the surface of the FPC connector to be inspected identified in the example shown in FIG. 5 .

Detailed ways

The specific embodiments of the present invention will be further described below with reference to the accompanying drawings.

The present application discloses a method for detecting surface cracks of FPC connectors based on machine vision. The method includes the following steps, please refer to FIG. 1 :

Step 110: Obtain the to-be-processed image F(x, y) of the to-be-detected area of the to-be-detected FPC connector. The area to be inspected may be the entire area of the FPC connector to be inspected, or the area of a part of the structure of the FPC connector to be inspected, which is determined according to the area where the surface crack inspection is actually to be performed. According to the actual situation, FPC connectors generally have surface cracks at the plastic end face, so the surface crack detection can also focus on this part, and the area to be detected is the end face area of the FPC connector to be detected. This application uses this as an example for follow-up instruction of.

The method for obtaining the image to be processed F(x,y) includes the following steps, please refer to the example diagram of FIG. 2:

(1) Obtain the original image of the region of interest of the FPC connector to be detected. The region of interest includes the region to be detected and other regions, usually the entire region of the FPC connector to be detected, because it is difficult to directly obtain an accurate to-be-processed image F(x,y) of the region to be detected in actual operation.

When actually acquiring the original image, the tooling slot is usually used to clamp the FPC connector to be tested for shooting, and it is usually difficult to directly acquire the original image of the region of interest. As shown in (a) of Figure 2, the full-area original image contains the region of interest and the background area, that is, the image containing the FPC connector to be tested, the tooling slot and other background images, and then the full-area original image can be obtained from the original image. The original image of the region of interest is extracted from the original image, as shown in (b) of Figure 2, the method of extracting the region of interest is not repeated in this application. This application uses a 5.1 million-pixel industrial microscope camera with USB3.0 interface for image acquisition. This camera can meet the high-precision imaging of micron-scale cracks on the surface of the FPC connector in this example. At the same time, to ensure stable imaging effect, a white LED ring The light source fills in light.

The extracted original image of the region of interest of the FPC connector to be detected contains the area between the upper and lower boundaries of the FPC connector to be detected in the vertical column direction, and the left boundary of the region of interest in the horizontal row direction is the FPC to be detected. The first metal pin of the connector and the right border are the last metal pin of the FPC connector to be detected;

(2) Process the original image of the region of interest into a binary image and determine the coordinates of the lower boundary pixel points. In order to reduce the amount of data processing, the lower half of the original image of the region of interest can be roughly intercepted, as shown in Figure 2 shown in (c). Then, the original image of the lower half of the region of interest is processed into a binary image, as shown in (d) in FIG. 2 . For the sake of clarity, this embodiment also performs color inversion on (d) to obtain (e), Then extract the coordinates of the location of the lower boundary pixel point.

The lower boundary pixel point is a pixel point in the binary image that is located at the lower boundary of the FPC connector to be detected and whose pixel value is different from that of the adjacent pixel point. Specifically: Convert the original image of the region of interest with RGB three channels to the image of the HSV three-pass, and extract the V channel as the preprocessed grayscale image.

Threshold segmentation is performed on the preprocessed grayscale image to obtain a binary image, and the segmentation threshold can be taken as 0.5. And extract the coordinates of the position of the lower boundary pixel where the pixel value of the adjacent pixel in the lower half of the binary image is abruptly changed.

(3) Perform straight line fitting based on the coordinates of the lower boundary pixel points in the binary image to obtain the lower boundary fitting line. specific:

The coordinates of any two lower boundary pixel points are fitted with a straight line to obtain a candidate fitting line, and the number of lower boundary pixel points whose distance from the candidate fitting line is within the error range is determined as the fitting line of the candidate fitting line. According to the exact number of data points, the error range can be customized to avoid the influence of the noise of the coordinates.

Re-execute the steps of performing straight line fitting on the coordinates of two optional lower boundary pixel points to obtain candidate fitting lines, until the iteration termination condition is reached, and the number of accurate fitting data points of each candidate fitting line is obtained. The iteration termination condition can be iterative The count reaches the count threshold. The candidate fitting line with the largest number of corresponding fitting accurate data points among the candidate fitting lines obtained in each iteration is used as the lower boundary fitting line.

(4) Translate the lower boundary fitting line upward according to the width d of the area to be detected to obtain the upper boundary fitting line. The width d is predetermined based on the size of the FPC connector to be inspected.

(5) Intercept the image between the upper boundary fitting line and the lower boundary fitting line from the original image of the region of interest to obtain the to-be-processed image, and then extract the to-be-processed image F(x,y) of the to-be-detected area. , in the example of this application, that is, the image of the end face area of the FPC connector to be detected. As shown in Fig. 2, the mask as shown in Fig. 2 (f) formed by the upper boundary fitting line and the lower boundary fitting line can be extracted from (b) to obtain the to-be-processed image F(x) of the area to be detected. , y), as shown in (g) in Figure 2.

Step 120 , using the structural elements S(x,y) of various typical surface crack features to perform multi-scale morphological operations on the to-be-processed image F(x,y) to obtain a two-dimensional crack probability map P(x,y).

The structural elements S(x, y) of various typical surface crack features are pre-established according to several typical surface crack features common to FPC connectors, and there are six typical ones in practical applications, as shown in Figure 3. The two-dimensional crack probability map P(x, y) is obtained as:

P(x,y)=max{(F·S _i )(x,y)}-F(x,y);

表示膨胀运算的算子，Θ表示腐蚀运算的算子，具体的计算过程本申请不详细展开。Among them, S _i (x,y) is the structuring element of the i-th typical surface crack feature, (F·S _i )(x,y)-F(x,y) represents the gray-level black hat operation, (F· S _i )(x,y) represents the gray level closing operation and

represents an operator of dilation operation, Θ represents an operator of erosion operation, and the specific calculation process is not detailed in this application.

Step 130, using the segmentation threshold t to perform global adaptive threshold segmentation on the two-dimensional crack probability map P(x, y) to obtain the crack binary map Q(x, y).

Please refer to Figure 4, the segmentation threshold t of this step is calculated by the following method:

(1) According to the fact that the actual structure of the crack is a vertical crack, the two-dimensional crack probability map P(x, y) is projected and transformed in the longitudinal column direction to obtain a one-dimensional crack probability map P'(x, y);

(2) Perform Gaussian filtering on the one-dimensional crack probability map P'(x, y) according to the preset interval h, and determine the central pixel column corresponding to the maximum value in the filtering result. The preset interval h is based on the approximate crack width and crack probability. The inclination range that appears is preset, for example, h=15.

(3) Extract the local area f(x,y) in the two-dimensional crack probability map P(x,y) that contains several pixel columns on both sides with the center pixel column in the center, and the center contained in the local area f(x,y) The number of the left and right pixel rows of the pixel row is preset, for example, the local area f(x, y) includes an area extending 10 pixel rows to the left and right from the center of the central pixel row.

(4) Use the maximum inter-class variance method to solve the segmentation threshold t in the local area f(x, y).

Then, the global two-dimensional crack probability map P(x, y) can be globally adaptively segmented by using the segmentation threshold t obtained by solving the local area f(x, y) to obtain Q(x, y), the two-dimensional crack probability map P(x,y) is shown in Figure 5, and Q(x,y) obtained by global adaptive threshold segmentation is shown in Figure 6.

In step 140, the connected domain search mark is performed on the crack binary image Q(x, y). In one embodiment, an 8-connection mode is adopted. Determine the target connected domain in which the preset connected domain characteristics are satisfied, and determine the response intensity value mean of the candidate crack region at the position of the target connected domain in the two-dimensional crack probability map P(x,y).

When the fitting path length l of a connected domain is greater than the length threshold, and the area reaches the area threshold, and the included angle between the main direction and the horizontal row direction reaches the angle threshold, it is determined that the connected domain is a target connected domain that satisfies the preset connected domain characteristics. In one embodiment, the length threshold is 15 pixels, the area threshold is 20, and the angle threshold is 50°.

In the process of searching the marked connected domain, the position of the upper end point, the position of the lower end point of the target connected domain, the first distance _Lu between the upper end point and the upper boundary of the area to be detected, the lower end point and the lower boundary of the area to be detected can be determined The second distance between L _d and other position identification information, based on these position identification information, the candidate crack region at the position of the target connected domain in P(x, y) can be located. Then, the grayscale average value of the pixels contained in the candidate crack region is taken as the response intensity value mean of the candidate crack region.

Step 150: If the size and position information of the target connected domain and the response intensity value mean of the corresponding candidate crack region both satisfy the corresponding preset conditions, it is determined that there is a crack on the surface of the FPC connector to be detected, otherwise it is determined that the FPC connector to be detected is cracked. There are no cracks on the surface.

In one embodiment, the size and position information of the target connected domain includes the fitting path length l of the target connected domain, the first distance Lu between the upper end point of the target connected domain and the upper boundary of the to-be _- detected area, the length of the target connected domain The second distance L _d between the lower end point and the lower boundary of the area to be detected.

和/或，

时，确定待检测FPC连接器的表面存在裂纹，LU是第一距离对应的距离阈值，LD是第二距离对应的距离阈值。Then, when l>L and at least one of L _u and L _d reaches the corresponding distance threshold, it is determined that the size and position information of the target connected domain satisfies the corresponding preset condition. When mean>M, it is determined that the response intensity value mean of the candidate crack region satisfies the corresponding preset condition, L is the fitting path length threshold, and M is the response intensity threshold. That is to say, when satisfied

and / or,

, it is determined that there is a crack on the surface of the FPC connector to be detected, LU is the distance threshold corresponding to the first distance, and LD is the distance threshold corresponding to the second distance.

In addition to determining that there is a crack on the surface of the FPC connector to be tested, the candidate crack area can also be identified by an index to indicate the location and shape of the crack, as shown in Figure 7.

The above descriptions are only preferred embodiments of the present application, and the present invention is not limited to the above embodiments. It can be understood that other improvements and changes directly derived or thought of by those skilled in the art without departing from the spirit and concept of the present invention should be considered to be included within the protection scope of the present invention.

Claims (10)

1. A method for detecting cracks on the surface of an FPC connector based on machine vision is characterized by comprising the following steps:

acquiring an image F (x, y) to be processed of a region to be detected of the FPC connector to be detected;

carrying out multi-scale morphological operation on the image F (x, y) to be processed by utilizing structural elements S (x, y) of various typical surface crack characteristics to obtain a two-dimensional crack probability map P (x, y);

carrying out global self-adaptive threshold segmentation on the two-dimensional crack probability map P (x, y) by utilizing a segmentation threshold t to obtain a crack binary map Q (x, y);

conducting connected domain search marking on the crack binary image Q (x, y), determining a target connected domain which meets the characteristics of a preset connected domain, and determining a response intensity value mean of a candidate crack region at the position of the target connected domain in the two-dimensional crack probability image P (x, y);

and if the size position information of the target connected domain and the response intensity value mean of the candidate crack region corresponding to the target connected domain both meet the corresponding preset conditions, determining that cracks exist on the surface of the FPC connector to be detected.

2. The method according to claim 1, wherein the acquiring of the image to be processed of the area to be detected of the FPC connector to be detected comprises:

acquiring an original image of an interested area of the FPC connector to be detected, wherein the interested area comprises an area between the upper boundary and the lower boundary of the FPC connector to be detected in the longitudinal column direction, the left boundary of the interested area in the transverse row direction is a first metal pin of the FPC connector to be detected, and the right boundary of the interested area is a last metal pin of the FPC connector to be detected;

processing the original image of the interested area into a binary image and determining the coordinates of the positions of lower boundary pixel points, wherein the lower boundary pixel points are pixel points which are positioned at the lower boundary of the FPC connector to be detected and have different pixel values from the adjacent pixel points in the binary image;

performing straight line fitting based on coordinates of lower boundary pixel points in the binary image to obtain a lower boundary fitting line, and translating the lower boundary fitting line upwards according to the width d of the area to be detected to obtain an upper boundary fitting line;

and intercepting an image between the upper boundary fitting line and the lower boundary fitting line from the original image of the region of interest to obtain the image to be processed.

3. The method according to claim 2, wherein performing straight line fitting based on coordinates of lower boundary pixel points in the binary image to obtain a lower boundary fitting line comprises:

carrying out linear fitting on the coordinates of optional two lower boundary pixel points to obtain a candidate fitting line, and determining the number of the lower boundary pixel points with the distance between the lower boundary pixel points and the candidate fitting line within an error range as the fitting accurate data point number of the candidate fitting line;

and re-executing the step of performing linear fitting on the coordinates of the optional two lower boundary pixel points to obtain a candidate fit line until an iteration termination condition is reached, and taking the candidate fit line with the largest number of accurate fitting data points in the candidate fit lines obtained by each iteration as the lower boundary fit line.

4. The method of claim 2, wherein processing the original image of the region of interest into a binary image and determining coordinates of locations of lower boundary pixel points comprises:

converting the original image of the region of interest of the RGB three channels into an image of the HSV three-way channel, and extracting a V channel as a preprocessing gray image;

and performing threshold segmentation on the preprocessed gray level image to obtain the binary image, and extracting coordinates of the position of a lower boundary pixel point where the pixel value of the lower half subregion of the binary image is mutated with the pixel value of an adjacent pixel point.

5. The method according to claim 1, wherein the multi-scale morphological operation of the image F (x, y) to be processed by using the structural elements S (x, y) of various typical surface crack features is to obtain a two-dimensional crack probability map P (x, y), and the two-dimensional crack probability map P (x, y) is obtained by:

P(x,y)＝max{(F·S_i)(x,y)}-F(x,y)；

wherein S is_i(x, y) is a structural element of the ith typical surface crack characteristic, (F.S)_i) (x, y) -F (x, y) represents the gray level black cap operation, (F.S)_i) (x, y) represents a gray level closure operation and

6. the method of claim 1, further comprising:

performing projection conversion on the two-dimensional crack probability map P (x, y) according to a longitudinal column direction to obtain a one-dimensional crack probability map P' (x, y);

performing Gaussian filtering on the one-dimensional crack probability map P' (x, y) according to a preset interval h, and determining a central pixel column corresponding to a maximum value in a filtering result;

extracting a local area f (x, y) which takes the central pixel column as the center and comprises a plurality of pixel columns at two sides in the two-dimensional crack probability map P (x, y);

and solving the segmentation threshold t in the local area f (x, y) by using a maximum inter-class variance method.

7. The method according to claim 1, wherein a connected domain is determined to be a target connected domain satisfying the preset connected domain characteristic when a fitting path length l of the connected domain is greater than a length threshold, an area reaches an area threshold, and an included angle between a main direction and a transverse row direction reaches an angle threshold.

8. The method according to claim 1, wherein the determining the response intensity values mean of the candidate crack regions at the position of the target connected domain in the two-dimensional crack probability map P (x, y) comprises:

and determining a candidate crack region at the target connected domain in the two-dimensional crack probability map P (x, y), and taking the gray average value of pixel points contained in the candidate crack region as the response intensity value mean of the candidate crack region.

9. The method of claim 1, wherein the size location information of the target connected components comprises a fit path of the target connected componentsThe length L of the diameter, and a first distance L between the upper end point of the target connected domain and the upper boundary of the area to be detected_uAnd a second distance L between the lower end of the target connected domain and the lower boundary of the area to be detected_d。

10. The method of claim 9,

when l is>L and L_uAnd L_dWhen at least one of the target connected domains reaches a corresponding distance threshold, determining that the size and position information of the target connected domain meets corresponding preset conditions; mean of>And M, determining that the response intensity value mean of the candidate crack region meets the corresponding preset condition, wherein L is a fitting path length threshold value, and M is a response intensity threshold value.

Images