JPH05107200A - Inspecting apparatus for internal surface of circular container - Google Patents

Abstract

(57) [Abstract] [Purpose] Conventionally, the method of differentiating the image density is used to detect white and black stains on the container surface.
This solves the problem that the annular high-intensity part appears in the image, the density change of the defect background is complicated, and the defect detection is difficult. [Structure] Gray-scale pixel value PO of a target point on the same scanning line
(I, j), and PO (i + α, j) with THD as a threshold value between the grayscale pixel values PO (i + α, j) and PO (i−α, j) of the background points that are separated by α pixels before and after that. -PO (i, j)> THD PO (i-α, j) -PO (i, j)> THD, the point of interest is a valley defect, and the relationship of the expression in which the sign of the above difference is inverted is In some cases, a combination of peak / valley detection means for making the point of interest a mountain defect and black / white level detection means for binarizing the grayscale image signal simply with upper and lower threshold values, and some target inspection areas And the .alpha. And THD are changed for each area, or the scanning direction in which the background frequency change is small is selected to detect the stain defect portion.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention inspects the inner surface of a circular container such as a beer can conveyed on a conveyor or the like, and
The present invention relates to a circular container inner surface inspection device as an image processing device for detecting dust, scratches, and the like. In the following figures, the same reference numerals indicate the same or corresponding parts.

[0002]

2. Description of the Related Art FIG. 8 is an explanatory view of a high-intensity part in the case of observing a container of an aluminum can for beer as an example, and FIG. 8A is a top view (image) of the container and FIG. B) is a side sectional view. 102 is a container, 101 is this container 1
A ring-shaped illuminator that illuminates 02 from above, 103 is a high-intensity part at the mouth, and 104 is a high-intensity part at the bottom. By thus irradiating the inner surface of the container 102 with the light beam using the ring illuminator 101, the mouth and the bottom of the inner surface of the container 1
High brightness areas such as 03 and 104 occur. This is particularly noticeable when the inner surface of the container has a metallic luster, such as a can.

FIG. 9B is a top view image of the container 102 (see FIG. 1).
3A shows changes in density on the scanning line Q-Q1 with respect to 3 (A)), and is classified into five regions W1 to W5 according to the characteristics of the density change. The first region W1 is the mouth high-intensity part 1
03, the second region W2 is the middle part on the side surface of the container where the change in concentration is relatively small, and the third region W3 is a darker container than the other regions because the light rays from the illumination 101 described in FIG. It is the lower part of the side surface, the fourth region W4 is the bottom high-intensity part 104, and the fifth region W5 is the bottom.

Conventionally, a window is provided in each of these areas W1 to W5, and a threshold value for detecting a defect of black stain (black spot) or white stain (white spot) is set according to the optical characteristics of the area. Was. As a method of detecting a defect, for example, an 8-bit multi-value grayscale image signal obtained by A / D conversion of an analog video signal (analog grayscale image signal) obtained by scanning a target image is set at a predetermined threshold value. A binarization method, a differentiation method of differentiating the video signal through a differentiation circuit as shown in FIG. 10 and extracting a defect signal, and the like are known. In the case of this differentiating method, a differential signal is also generated in the contour portion of the outer shape of the object, but in the contour portion, one of the positive direction pulse and the negative direction pulse is generated by differentiation, whereas in the minute defect portion, the positive direction pulse is generated. The defect can be extracted by utilizing the fact that the negative pulses are generated at the same time.

That is, a value P (i, j) at a target point (coordinate value x = i, y = j) of a signal P (x, y) obtained by differentiating an analog grayscale image signal based on raster scanning,
The value P (i- at a point separated by a predetermined minute α pixel and β pixel from the point of interest on the x-direction scanning line, respectively.
between α, j) and P (i + β, j), P (i, j) −P (i−α, j)> TH1 and P (i + β, j) −P (i, j) )> TH1 (if TH1 is a predetermined threshold value (positive value)) Binarization function value PD for defect detection at the point of interest
When (i, j) = 1, this point of interest is a black level defective point, and in other cases, PD (i, j) = 0 is set and this point of interest is a normal point.

[0006]

FIG. 11 is an explanatory view of the problems of the conventional defect detection method based on the differential method.
Shows an example of a density change (analog grayscale image signal) on the scanning line Q-Q1, FIG. 7B shows an analog differential signal corresponding to FIG. 4A, and FIG. Examples of digital differential signals corresponding to BD in each of these figures (A) to (C) is a black stain (valley) defective portion.
That is, in the conventional defect detection method, even if a black level defect such as BD exists as a signal in a portion where the density change is inclined as shown in FIG. 11A, according to the analog differentiation method, FIG. As shown in Fig. 5, the time constant of the filter circuit only causes the differential signal of the small defective portion to be superimposed on the grayscale differential signal that is the base, and the digital differential method does not stabilize the signal as shown in Fig. 7C. The differential signal of the defective portion is buried in the noise component, and it is extremely difficult to detect the defective signal using a certain threshold value.

FIG. 12 shows an example of a top view of the container 102 having an angular protrusion 102a on the bottom, but on the other hand,
As described above, the inner surface of the container has multiple high-intensity parts as shown in FIGS. 104-1 and 104-2 depending on the shape of the bottom and the degree of reflection of light rays on the side surface. Is often mirror-like, and this tendency is remarkable. It is difficult to remove such a high-intensity part by devising lighting, and when inspecting the inner surface of the container, it is essentially necessary to adopt a defect detection method that corresponds to the uneven illuminance on the inner surface of the container by the high-intensity part. However, it has been difficult to detect this defect by the conventional method. Therefore, it is an object of the present invention to provide a circular container inner surface inspection apparatus capable of detecting a defective portion with stability and high accuracy even when the inner surface of the container has uneven illuminance.

[0008]

In order to solve the above-mentioned problems, an apparatus for inspecting an inner surface of a circular container according to a first aspect of the present invention is arranged such that an axially symmetric circular container (102, etc.) is provided with a ring illuminator 101, etc. After illuminating the inner surface side of the circular container, the illumination surface of the circular container is imaged from the axial direction through a TV camera, and the captured image is analyzed to black and white the inner surface of the circular container. In a circular container inner surface inspection device for inspecting stains, a grayscale image signal (inspection region grayscale image signal 23a or the like) obtained by screen scanning of the image pickup.
Of the value (PO (i, j), etc.) of the pixel of interest on the same screen scanning line for the background is separated by a predetermined number (α, etc.) of pixels (hereinafter referred to as α pixels) before and after this pixel of interest. Pixel values (hereinafter referred to as front background pixel and rear background pixel, respectively) (PO (i + α, j), PO (i−α,
j) etc.) is the same polarity, and the absolute value of a predetermined one of the two differences is greater than a predetermined first threshold value (THD, etc.) corresponding to said polarity, and When the other absolute value of the difference is larger than a predetermined second threshold value (THD or the like) corresponding to the polarity, the peak / valley defect determining means (peak / valley detection binary value) for determining the pixel of interest as defective. Of the AND gate 39 in the digitizing circuit 24) and the optical characteristics of the inner surface of the circular container based on the illumination, the inspection area of the image (Z1 to Z4) that is the target of the peak / valley defect determining means. , Za to Zc), and a means for varying at least one of the number of α pixels, a first threshold value, and a second threshold value for each of the divided inspection areas. Assume that

A circular container inner surface inspection apparatus according to a second aspect is the circular container inner surface inspection apparatus according to the first aspect, in which the number of the α pixels is changed for one of the inspection areas, and the peak / valley defect determining means is provided. Provided with means for repeating the process,

A circular container inner surface inspection apparatus according to a third aspect is the circular container inner surface inspection apparatus according to the first or second aspect, in which the value of the grayscale image signal is predetermined for each of the inspection regions. Threshold (black level threshold THB
Etc.) and a black level defect determining means (comparator 37-3 etc.) for determining smaller pixels as defective,

According to a fourth aspect of the present invention, there is provided a circular container inner surface inspection apparatus according to any one of the first to third aspects, wherein the grayscale image signal value is predetermined for each of the inspection regions. Threshold (white level threshold THW
Etc.) and a white level defect judging means (comparator 37-4 etc.) for judging a larger pixel to be defective,

A circular container inner surface inspection apparatus according to a fifth aspect is the circular container inner surface inspection apparatus according to the first aspect, wherein a 1 or 2's complement of the grayscale image signal is obtained to obtain an inverted grayscale image signal, and the signal is obtained. In place of the grayscale image signal, means for supplying the peak / valley defect determining means is provided, and the pixels with peak / valley defects are detected without reversing the polarity,

A circular container inner surface inspection apparatus according to a sixth aspect of the present invention is the circular container inner surface inspection apparatus according to the third aspect, wherein a 1 or 2's complement of the grayscale image signal is obtained to obtain an inverted grayscale image signal, and the signal is obtained. Is provided to the black level defect determining means instead of the grayscale image signal, and a pixel having a white level defect is also detected by using the black level defect determining means.

A circular container inner surface inspection device according to claim 7 is the circular container content inspection device according to any one of claims 1 to 6, wherein the scanning direction of the image is horizontal, vertical, or slanted for each inspection region. And a means for selecting a direction (AR or the like) in which the change of the grayscale image signal has the lowest frequency among a plurality of predetermined directions such as a direction.

The circular container inner surface inspection device according to claim 8 is the circular container inner surface inspection device according to any one of claims 1 to 7, wherein a predetermined value including the pixel on the screen scanning line is included as a value of the front background pixel. Means (smoothing circuit 41-1 or the like) for outputting an average value of pixel values in a section consisting of the first number of pixels (n or the like), and the pixel on the screen scanning line as the value of the rear background pixel. And a means (smoothing circuit 41-2 or the like) for outputting an average value of pixel values in a section including a predetermined second pixel number (n or the like) including

A circular container inner surface inspection apparatus according to a ninth aspect is the circular container inner surface inspection apparatus according to the eighth aspect, wherein the two means are replaced by the median of pixel values of corresponding sections instead of the average value. It shall be a means to do.

[0017]

Operation: Multi-value grayscale image signal PO based on raster scanning
Point of interest for (x, y) (coordinate values x = i, y = j)
Value (pixel value of interest) PO (i, j) and a value (background pixel value) at two points (background points) that are apart from this point of interest on the x-direction scanning line by a predetermined small α pixel, respectively.
PO (i + α, j) and PO (i−α, j) are extracted,
The relationship between these three points forms a valley shape in the black level detection and a mountain shape in the white level detection (that is, the density difference between the background pixel value and the pixel value of interest), and the absolute value of the density difference is detected. When the pixel value exceeds a certain threshold value THD, the pixel value of interest PO (i, j) is regarded as a defective pixel.

That is, FIG. 1 is an explanatory view of the principle of the valley detection binarization method which is the core of the present invention, and FIG. 1A shows a scanning line (y = j).
An example of the multi-value grayscale image signal PO (x, y) on Q-Q1 is shown. Here, 51 is the point of interest in the non-defective part, and 52 and 53.
Are the background points of the non-defective part and the background points of the non-defective part as background points separated by the number of pixels α in the front and rear of the point of interest 51, respectively. Similarly, 54 is a target point in the defective portion, and 55 and 56 are a defective portion front background point and a defective portion rear background point as background points which are separated from the focused point 54 by the number of pixels α before and after on the scanning line. ..

In the present invention, the coordinates of the point of interest are x = i, y = j
Then, PO (i-α, j) -PO (i, j)> THD (1) and PO (i + α, j) -PO (i, j)> THD (2) ) (However, if THD has a predetermined threshold value (positive value)), a binarization function value (referred to as a peak / valley defect binarized image signal) for defect detection at a point of interest POD ( i,
j) = 1, this point of interest is taken as a valley (defective). In the non-defective part of FIG. 1, the above equation (2) is not satisfied and no defect is detected, but in the defective part of FIG. The above (1) and (2) are established, and the valley defect can be detected. Note that FIG. 1 (B) corresponds to FIG. 1 (A), and the peak / valley binarized image signal POD (x, j) as a defect determination output.
Indicates.

In this way, the waveform of FIG. 1A is divided into small regions, and the number of pixels α and the threshold value THD in the equations (1) and (2) are given to each of these small regions as appropriate. Optimal detection performance can be obtained. Further, according to the present invention, in the case of detecting a peak (defective), the positions of the difference terms in the expressions (1) and (2) are reversed, and PO (i, j) -PO (i-α, j)> THD ... If (1A) and there is a relationship of PO (i, j) -PO (i + α, j)> THD (2A), the peak / valley binarized image signal POD (i, j) = 1, and this point of interest is regarded as a mountain defect.

[0021]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENT An embodiment of the present invention will be described below with reference to FIGS. FIG. 2 is a block diagram of hardware as an embodiment of the present invention. In the figure, PO is a multi-valued (e.g. 8-bit) grayscale image signal obtained by AD-converting a video signal obtained by raster scanning the surface of a TV camera (not shown), and 1 is this multivalued grayscale image signal PO. Is input and stored as multi-valued screen data. Reference numeral 3 is an address generation circuit for this frame memory. Reference numeral 2 is a window memory in which a mask pattern for each window is stored, and 4 is an address generation circuit for this window memory. A window gate circuit 5E masks the multi-value grayscale image signal PO or the image signal 1a read from the frame memory 1 with the mask pattern data 2a from the window memory 2 to generate the image signal PO or 1a only in the designated window area. It is a window gate circuit that passes through.

Image edge detection circuits 6-1 and 6-2 are
It has a function to detect the edge of the image, specifically the outer edge (outer peripheral point) and inner edge (inner peripheral point) of the ring-shaped high-intensity part. In this case, the input image signal is used to detect the position of the target image and After binarizing with a predetermined threshold value for the circularity inspection, the coordinates of the rising point and the coordinates of the falling point of this binarized signal as an image edge are stored in its own memory. Reference numeral 11 is a circuit for performing circularity inspection on the coordinate values of the outer peripheral points or the inner peripheral points detected by the image edge detection circuit 6-1. 13 is preset with the center position of the actual target image detected by the image edge detection circuit 6-2 by inputting the latest multi-value grayscale image signal PO so that a window is generated at the correct position for the target image. This is a circuit for detecting a deviation from the center position of the window being displayed.

Reference numeral 21 denotes a defect inspection target area for each horizontal scanning line (in other words, for the target container) in order to input the multi-value grayscale image signal 1a which has passed through the window gate circuit 5E from the frame memory 1 and detect defective pixels. A region detection circuit that outputs a region signal 21a as a signal that defines a region divided by the outer shape, 22 is synchronized with this region detection circuit 21, and the same image signal 1a for each one horizontal scanning line is input and temporarily stored. 23 is a line memory that performs an AND condition between the area signal 21a and the grayscale image signal 22a output from the line memory 22 as an image signal for each horizontal scanning line corresponding to the area signal 21a. It is an AND gate that outputs a grayscale image signal (referred to as an inspection region grayscale image signal) 23a of only the region. Reference numeral 24 is a peak / valley detection binarization circuit for detecting defective pixels including the peak / valley defects described in FIG. 1 from the image signal 23a, which is the core of the present invention.

Next, 9 is an X projection circuit for obtaining a projection pattern in the X direction of the target image using the multi-valued image signal PO which has passed through the window gate circuit 5E, and 10 is also Y of the target image.
A Y projection circuit for obtaining the directional projection pattern, and a circuit 14 for obtaining only the region of the target container image which is not connected to other container images by using the output data of the two projection circuits 9 and 10. Further, 15 is a circuit for inputting the judgment results of the high-luminance part judgment circuit 11 and the peak / valley detection binarization circuit 24 to make a comprehensive judgment, and 16 is a pass / fail output according to the output judgment signal of this comprehensive judgment circuit 15. It is an output circuit.

FIG. 3 is a block diagram showing an embodiment of the detailed configuration of the peak / valley detection binarization circuit 24 of FIG. However, this configuration shows the case of valley (defective) detection. In the case of peak (defective) detection, the polarity of the subtraction of the subtraction circuits 36-1 and 36-2, which will be described later, is reversed or the peak / valley detection is performed. The input image signal 23a to the binarization circuit 24 is inverted. In this latter case, the function of the comparator 37-3 (black level determination by fixed binarization) and the function of the comparator 37-4 (white level determination by fixed binarization) are switched.

Next, the function of FIG. 3 will be described. This Figure 3
Implements the principle of FIG. 32 in FIG.
-1, 32-2 are input image signals (that is, A described in FIG. 2).
Inspection area grayscale image signal as output of ND gate 23)
23a is a + α pixel delay circuit that sequentially delays α pixels in the scanning direction. Reference numerals 41-1 and 41-2 are smoothing circuits for smoothing the image signal as necessary to reduce the influence of noise as described later, and 42 is a smoothing circuit ON for switching whether or not to perform this smoothing. / Off switch. Here, the smoothing circuit 41-1 is provided corresponding to the front background point detection circuit 33 described later, and the smoothing circuit 41-1 is also provided.
-2 is provided corresponding to the rear background point detection circuit 35 which will be described later. Note that there is no smoothing circuit corresponding to the point-of-interest detection circuit 34, which will be described later, because the point-of-interest defect detection sensitivity (that is, low peaks / valleys, light shades, in other words, defective pixels detected with a small threshold value are detected. This is to improve the ability to do).

The front background point detection circuit 33 receives the inspection area grayscale image signal 23a as the original input image signal or its smoothing signal to detect the front background point, and the target point detection circuit 34 is + α pixel. The circuit for detecting the point of interest by inputting the output image signal of the delay circuit 32-1 and the rear background point detecting circuit 35 for inputting the output image signal of the + α pixel delay circuit 32-2 or the smoothed signal thereof to the rear background point. Is a circuit for detecting the
When the detection circuits 33, 34, and 35 do not use the smoothing circuits 41-1 and 41-2 by the action of 2-2 (that is, the smoothing circuits are short-circuited by the switch 42), the pixels described in FIG. Values PO (i + α, j), PO (i, j), P
Latch O (i-α, j) simultaneously.

However, when the smoothing circuits 41-1 and 41-2 are used, the pixel values PO (i + α, j) and PO (i-
α and j) are replaced with the values of the following expressions (3) and (4), respectively. PO (i + α, j) = { _{k = 0} Σ ^n-1 PO (i + α + k, j)} / n (3) PO (i-α, j) = { _{k = 0} Σ ^n-1 PO (i-α -K, j)} / n (4) That is, the smoothing circuit 41-1 determines the pixel value P of the corresponding front background point.
The average of n pixel values on the front side including O (i + α, j) is calculated and replaced with the pixel value of the front background point. Similarly, the smoothing circuit 41-2 is used for the rear background point. Of the pixel value PO (i-α, j), the total of n pixel values on the rear side is calculated and replaced with the pixel value of the rear background point. It should be noted that the reason why the average value centering on the pixel value is not calculated here is to prevent the pixel value PO (i, j) of the point of interest from being involved in the calculation of this average value. Further, instead of obtaining the average value as in the above equations (3) and (4), the median of the n pixel values (that is, n
It is also possible to extract a value positioned in the central order when the individual values are arranged in the order of magnitude).

Now, the above-mentioned respective image data latched by the detection circuits 33, 34 and 35 of FIG. 3 are subtracted by the subtraction circuits 36-1 and 36-3.
6-2, which are described in FIG. 1 (1),
The difference is calculated according to the contents of the equation (2). This difference is the peak / valley threshold value T set in the threshold value setting circuits 38-1 and 38-2 by the comparators 37-1 and 37-2, respectively.
Compared with HD, the AN of these comparators 37-1, 37-2
As the output of the AND gate 39 for obtaining the D condition, the peak / valley binary image signal PO described in FIG.
D is obtained. On the other hand, the comparators 37-3 and 37-4 are for detecting a defective pixel having a relatively large area, and the comparator 37-3 outputs the image data of the pixel of interest from the point-of-interest detection circuit 34 which does not input the smoothed image signal. The output is received and compared with the black level threshold THB set in the threshold setting circuit 38-3, and the black level binarized image signal 37B indicating the black level defective pixel is detected and output. Similarly, the comparator 37-4 receives the pixel data output of the pixel of interest and receives the threshold value setting circuit 3
Compared with the white level threshold THW set to 8-4,
White level binarized image signal 37W indicating a white level defective pixel
Is detected and output.

The OR gate 40 is provided with the peak / valley binary image signal P as each defective pixel detection signal thus detected.
Means for obtaining an OR condition of OD, black level binary image signal 37B, white level binary image signal 37W, and outputting a defective binary image signal 40a to detect a black level defective pixel (comparator 37-3. Etc.) and a means (comparator 37-4 etc.) for detecting a white level defective pixel are processed in parallel with the peak / valley detection binarization means (AND gate 39 etc.) in the example of FIG. Of course, these means are included in the scope of the present invention even if a method of individually inspecting an image, finally synthesizing each operation, and making a judgment output is used.

Next, the image scanning method will be described. Figure 1
(A) shows the light and shade of the image in a certain cross section Q-Q1 of the inner surface of the container. The number of pixels α in the formulas (1) and (2) or the formulas (1A) and (2A) is a defective portion. It is a parameter that gives the frequency of the image signal of a part (in other words, the width of a mountain or valley). However, as shown in FIG. 1 (A), the grayscale image signal on the inner surface of the container at the time of non-defective product is complicated including many kinds of frequency components, and as described above, the inner surface of the container is divided and optimum parameters are set for each. Need to give. However, by devising the scanning direction of the image, it is possible to further reduce the grayscale change of the background pixel and improve the detection accuracy.

FIG. 4 is an explanatory view of an embodiment of such an image scanning method. In FIG. 4A, WB is a window for selecting a defect detection target area, Z1, Z2, Z3.
Z4 is an upper circle area, a lower circle area, a left circle area, and a right circle area, respectively. That is, in FIG. 4, the bottom high-intensity portion 104 in which the shade of the container 102 changes a lot is selected in the window WB, and the valley detection binarization is tried. Here, for example, when the shade change in the horizontal scanning direction in the area of the window WB is examined, a portion where the shade changes at a high frequency, such as HF in FIG. 4C, occurs in the left circular area Z3, which affects the detection sensitivity. But give
When the left circular area Z3 of FIG. 4A is scanned in the direction of the scanning direction arrow AR, a low-frequency gradation change is obtained as a background as shown in FIG. Since the frequency is extremely high, the accuracy of detection can be improved.

FIG. 5 is an explanatory view of an embodiment of a simple image scanning method. In the figure, the scanning direction remains horizontal as indicated by the arrow AR, and the scanning area is the window WB.
Is divided into three areas, Za, Zb, and Zc. In this case, threshold values for detecting peaks and valleys in the areas Za and Zc and threshold values for detecting peaks and valleys in the area Zb are set. The setting is performed separately, and the optimum detection is performed according to the frequency of the light and shade of the background. In this case, although the defect detection sensitivity in the area Zb is low, the detection sensitivity in the areas Za and Zc can be increased.

FIG. 6 shows a modification of FIG. 4, in which the scanning area is equally divided into four fan-shaped areas around the center of the container, and in FIG. 6, it is equally divided into eight fan-shaped areas. The optimum scanning direction AR is given to the area.
The detection sensitivity can be higher than in the case of.

FIG. 7 is an explanatory diagram of the relationship between the defect detection method according to the present invention and the shape of the defective portion. In the same figure (A), 71-73 show the defective part which appeared in the image of the container 102. The same figure (B) shows the shade change in the scanning cross section QA-QA1 of the same figure (A), but the oval defective portions 71, 72.
The frequency of light and shade changes depending on the direction of the long (short) diameter. Therefore, in the same inspection area, the above equation (1),
The defect detection capability can be improved by selecting some kinds of amounts corresponding to the pixel number α in (2) and repeating the inspection.

On the other hand, FIG. 7C shows a scanning section Q of FIG. 7A.
It is assumed that the change in shade in B-QB1 is large, but the defective portion 73 on this cross section is large, and the change in the shade image signal has a low frequency but is sufficiently black with respect to the background. In this case, since the defective portion 73 has a low frequency, it cannot be detected in the valley detection binarization unless the value of the number of pixels α is extremely large, which is not practical. In such a case, the black level threshold value THB is set in the threshold value setting circuit 38-3 in FIG. 3 by using the fixed binarization method described in FIG.
As described above, the defective portion 73 is separated and detected as a black level to supplement the detection capability of the valley detection binarization. As a method for determining the black level threshold value THB, as an example, FIG.
An average of density data of pixels on a circumference (illuminance measurement circle) such as 74 in (A) is obtained, and a threshold THB is obtained by subtracting a set amount σ from this average value as shown in FIG. There is a method such as.

[0037]

According to the first aspect of the present invention, the inner surface of the circular container 102 is illuminated from the axial direction of the axisymmetric circular container 102 via the ring illuminator 101, and the axial direction is obtained from the TV camera. The illumination surface of the circular container 102 is imaged, the imaged image is analyzed, and the circular container 10 is analyzed.
In a circular container inner surface inspection device for inspecting black stains and white stains on the inner surface of No. 2, the value PO (i, j) of the pixel of interest on the same screen scanning line for the grayscale image signal 23a obtained by the screen scanning of the imaging is calculated. , A predetermined same number α of pixels before and after this pixel of interest (hereinafter referred to as α pixels)
Values PO (i + α, j), PO (i−α,) of background pixels (hereinafter referred to as front background pixel and rear background pixel, respectively) separated by
j), the two differences have the same polarity, and the absolute value of a predetermined one of the two differences is greater than a predetermined first threshold value (THD, etc.) corresponding to the polarity, and When the absolute value of the other of the two is larger than a predetermined second threshold value (THD, etc.) corresponding to the polarity, the peak / valley defect determining means (peak / valley detection binarization circuit) for determining the pixel of interest as defective. 24 and the optical characteristics of the inner surface of the circular container based on the illumination, the inspection area of the image to be the target of the peak / valley defect determination means (Z1 to Z).
4, Za to Zc) and a means for varying at least one of the number of α pixels, a first threshold value and a second threshold value for each of the divided inspection areas. Be prepared,

According to the invention of claim 2, in the circular container inner surface inspection device of claim 1, the number of the α pixels is changed for one of the inspection areas, and the processing of the peak / valley defect determining means is performed. It should be equipped with a means to repeat.

According to the third aspect of the invention, in the circular container inner surface inspection apparatus according to the first or second aspect, the value of the grayscale image signal 23a is set to a predetermined third value for each inspection area. Black level threshold TH as a threshold
A black level defect determining unit (comparator 37-3 or the like) for determining pixels smaller than B as defective is provided.

According to the fourth aspect of the present invention, in the circular container inner surface inspection apparatus according to the first to third aspects, the value of the grayscale image signal 23a is a predetermined fourth value determined for each of the inspection areas. White level threshold TH as a threshold
A white level defect determining unit (comparator 37-4 or the like) for determining pixels larger than W as defective is provided.

According to the invention of claim 5, in the circular container inner surface inspection device according to claim 1, the grayscale image signal 2
3a is obtained to obtain a 1 or 2's complement to obtain an inverted grayscale image signal, and means for supplying the signal to the peak / valley defect determining device in place of the grayscale image signal 23a is provided. So that the defective pixels are detected,

According to a sixth aspect of the present invention, in the circular container inner surface inspection apparatus according to the third aspect, the grayscale image signal 2
3a is obtained to obtain a 1 or 2's complement, an inverted grayscale image signal is obtained, and means for supplying the signal to the black level defectiveness judging means in place of the grayscale image signal 23a is provided, and the black level defectiveness judging means is used. Detect pixels with white level defects,

According to a seventh aspect of the present invention, in the circular container inner surface inspection apparatus according to the first to sixth aspects, a horizontal direction is set as a scanning direction of the image for each of the inspection areas.
A means for selecting a direction AR in which a change in the grayscale image signal has the lowest frequency among a plurality of predetermined directions such as a vertical direction and an oblique direction is provided.

According to the invention of claim 8, in the circular container inner surface inspection device according to any one of claims 1 to 7, a predetermined first value including the pixel on the screen scanning line concerned as a value of the front background pixel. A smoothing circuit 41-1 as a means for outputting an average value of pixel values in a section consisting of one pixel number (n or the like), and a predetermined value including the pixel on the screen scanning line as a value of the rear background pixel And a smoothing circuit 41-2 as a means for outputting the average value of the pixel values in the section consisting of the second number of pixels (n etc.),

According to the invention of claim 9, in the circular container inner surface inspection apparatus of claim 8, the two means are
As the means for outputting the median of the pixel value of the corresponding section instead of the average value, even if there is uneven illuminance on the inner surface of the circular container due to the high-intensity part or the like generated by illumination, the defective part can be accurately detected. can do.

[Brief description of drawings]

FIG. 1 is an explanatory view of the principle of a valley detection binarization method based on the present invention.

FIG. 2 is a block diagram showing a hardware configuration as an embodiment of the present invention.

FIG. 3 is a block diagram showing a detailed configuration of a peak / valley detection circuit as one embodiment of the present invention.

FIG. 4 is an explanatory diagram of a first embodiment of a screen scanning method according to the present invention.

FIG. 5 is an explanatory diagram of a second embodiment of the screen scanning method according to the present invention.

FIG. 6 is an explanatory diagram of a third embodiment of the screen scanning method according to the present invention.

FIG. 7 is an explanatory diagram of a relationship between a defect detection method according to the present invention and a shape of a defective portion.

FIG. 8 is a diagram showing a high-intensity part on the inner surface of a circular container.

FIG. 9 is a diagram showing a relationship between a change in concentration on the inner surface of a circular container and conventional window division.

FIG. 10 is a diagram showing an example of an analog differentiating circuit.

FIG. 11 is an explanatory diagram of a conventional defect detection method.

FIG. 12 is a diagram showing a high-intensity portion on the inner surface of a circular container when the bottom shape of the container is different.

[Explanation of symbols]

 PO multi-value grayscale image signal 1 frame memory 1a frame memory output image signal 2 window memory 2a mask pattern data 3 image address generation circuit 4 window address generation circuit 5E window gate circuit 6-1 image edge detection circuit 6-2 image edge detection circuit 9 X projection circuit 10 Y projection circuit 11 High-luminance part determination circuit 13 Position shift amount determination circuit 14 Processing area determination circuit 15 Overall determination circuit 16 Output circuit WB window 21 Area detection circuit 22 Line memory 23 AND gate 23a Inspection area grayscale image signal 24 mountain / valley detection binarization circuit 32-1 + α pixel delay circuit 32-2 + α pixel delay circuit 33 front background point detection circuit 34 target point detection circuit 35 rear background point detection circuit 36-1 subtraction circuit 36-2 subtraction circuit 37-1 Comparator 37 2 Comparator 37-3 Comparator 37-4 Comparator 37B Black level binarized image signal 37W White level binarized image signal 38-1 Threshold setting circuit 38-2 Threshold setting circuit 38-3 Threshold Value setting circuit 38-4 Threshold setting circuit 39 AND gate 40 OR gate 40a Defective binarized image signal 51 Non-defective part attention point 52 Non-defective part front background point 53 Non-defective part rear background point 54 Defective part attention point 55 Defective part front Background point 56 Backside point of defective part Z1 Upper circle area Z2 Lower circle area Z3 Left circle area Z4 Right circle area Za area Zb area Zc area 101 Ring illuminator 102 Container 103 Mouth high brightness part 104 Bottom high brightness part 104-1 Bottom high-intensity part 104-2 Bottom high-intensity part PO (i-α, j) Pixel value of rear background point PO (i, j) Pixel value of target point PO (i + α, j) Front spine Pixel value of scene point POD peak / valley binary image signal THD peak / valley threshold THB black level threshold THW white level threshold

Claims (9)

[Claims] 1. An inner surface side of this circular container is illuminated from the axial direction of an axially symmetric circular container, and an illumination surface of this circular container is imaged from this axial direction via a TV camera. In the circular container inner surface inspection device for inspecting the black stain and the white stain on the inner surface of the circular container by analyzing, the value of the pixel of interest on the same screen scanning line for the grayscale image signal obtained by the screen scanning of the imaging, Two differences subtracted from the values of background pixels (hereinafter, respectively referred to as front background pixel and rear background pixel) separated by a predetermined same number of pixels (hereinafter referred to as α pixels) before and after the target pixel have the same polarity, A predetermined absolute value of one of the two differences is larger than a predetermined first threshold value corresponding to the polarity, and a predetermined absolute value of the difference is a predetermined first threshold value corresponding to the polarity. From threshold A peak / valley defect determining unit that determines the pixel of interest as defective when the other absolute value of the difference is larger than a predetermined second threshold value corresponding to the polarity; and a circular container based on the illumination. Depending on the optical properties of the inner surface,
Means for dividing the inspection area of the image as the target of the peak / valley defect determining means, and at least the number of the α pixels, the first threshold value, and the second threshold value for each of the divided inspection areas. An apparatus for inspecting an inner surface of a circular container, which is provided with a means for changing one. 2. An apparatus for inspecting an inner surface of a circular container according to claim 1, further comprising means for changing the number of the α pixels in one of the inspection areas and repeating the processing of the peak / valley defect determining means. An inner surface inspection device for a circular container, which is characterized in that 3. The circular container inner surface inspection apparatus according to claim 1 or 2, wherein a pixel whose value of the grayscale image signal is smaller than a predetermined third threshold value determined for each inspection region is defective. An apparatus for inspecting an inner surface of a circular container, comprising: a black level defect determining means for determining 4. The circular container inner surface inspection apparatus according to claim 1, wherein a pixel whose value of the grayscale image signal is larger than a predetermined fourth threshold value determined for each inspection area is defective. An apparatus for inspecting an inner surface of a circular container, comprising: a white level defect determining means for determining 5. An apparatus for inspecting an inner surface of a circular container according to claim 1, wherein an inverse gray image signal is obtained by obtaining a 1 or 2's complement of the gray image signal, and the signal is replaced with the gray image signal. An apparatus for inspecting an inner surface of a circular container, which is provided with a means for supplying a defect determination device for peaks and valleys to detect pixels with defective peaks and valleys without reversing the polarity. 6. The circular container inner surface inspection apparatus according to claim 3, wherein a 1 or 2's complement of the grayscale image signal is obtained to obtain an inverted grayscale image signal, and the signal is replaced with the grayscale image signal. An apparatus for inspecting an inner surface of a circular container, comprising means for giving a black level defect judging means, wherein the black level defect judging means is also used to detect pixels having a white level defect. 7. The circular container inner surface inspection apparatus according to claim 1, wherein a predetermined plurality of directions such as a horizontal direction, a vertical direction, and an oblique direction are set as the scanning direction of the image for each of the inspection regions. An apparatus for inspecting an inner surface of a circular container, comprising means for selecting a direction in which a change in the grayscale image signal has the lowest frequency. 8. The circular container inner surface inspection device according to claim 1, wherein a section having a predetermined first number of pixels including the pixel on the screen scanning line concerned as a value of the front background pixel. Means for outputting the average value of the pixel values in, and means for outputting the average value of the pixel values in a section including a predetermined second number of pixels including the pixel on the screen scanning line as the value of the rear background pixel. An apparatus for inspecting the inner surface of a circular container, comprising: 9. The apparatus for inspecting the inner surface of a circular container according to claim 8, wherein the two means are means for outputting medians of pixel values of corresponding sections instead of the average value. Circular container inner surface inspection device.