JPH0242407B2 - - Google Patents

Description

[Detailed Description of the Invention] [Technical Field] The present invention captures an image of irregularities on the surface of an object to be inspected using a television camera, and recognizes the defects by performing multivalue numerical processing using a computer. The present invention relates to an uneven defect detection method.

[Background technology]

In the conventional unevenness defect detection method, first, a specularly reflected image obtained by applying specularly reflected light to an object to be inspected is processed by a binarization process such as floating binarization. Here, floating binarization compares input signals at a certain threshold,
When dividing into binary values (for example, 1, 0) based on magnitude, the threshold value changes according to the input signal. After that, they were put into a computer and used to count the number of pixels and use matching means to detect defects. That is, after binarizing the image,
A mask is prepared for the collection of images, and the original image is determined using this mask. In other words, defects were detected by overlapping a mask on the original image (matching) and counting the number of pixels in the area where the mask and the original image overlapped (pixel count). However, unevenness defects on the surface of the object to be inspected occur at random locations, and the matching method cannot be used. That is, the matching method cannot be used unless the object to which the method is applied is known, or in other words, the mask cannot be defined. Furthermore, since the uneven defects are not uniform in size, the method of counting the number of pixels has also become inaccurate.

[Purpose of the invention]

An object of the present invention is to enable processing using the same criteria even if the size of the defect is not uniform, thereby making it possible to detect uneven defects that are difficult to detect using conventional binarization processing.

[Disclosure of the invention]

Embodiment In FIG. 1, reference numeral 1 denotes a light source that illuminates an object 2 to be inspected obliquely from above and provides specularly reflected light to the surface of the object 2 to be inspected. A television camera 3 captures a regular reflection image of the object 2 to be inspected. At this time, the specular reflection image of the convex defect becomes a bright part 4 on the A side and a dark part 5 on the B side, as shown in FIG. TV camera 3
The video signal is input to the A/D converter 6, A/D converted, and sequentially written into the frame memory 7, as shown in FIG. 8 is a computer,
The image data of the inspected object 2 written in the frame memory 7 is numerically processed. 9 is a memory in which processing results and intermediate results are written. The frame memory 7 is a high-speed, large-capacity memory that can write image data for one image in synchronization with the television camera 3, and stores one image by distributing it into 256 x 256 pixels. It has 256 gradations (8 bits) of light intensity data.

Operation Divide the 256 x 256 pixels written in the frame memory 7 as shown in Figure 4 a into N x M areas by dividing them into N horizontally and M vertically as shown in Figure 4 b. . In the case of Figure 4b, M=8, N=8 and 64
The number of pixels in one area is 32.
×32=1024 pixels. in this one area
The computer 8 calculates the average light amount for 1024 pixels and writes it into the memory 9. This is calculated for all 64 areas, and the calculated average light quantity is assigned as K ₁₁ and K ₁₂ along the position of the area as shown in FIG. From now on, the average light intensity will indicate the position in K _MN . The reason why the frame memory 7 is divided into regions each containing a plurality of pixels and the average amount of light within each region is determined is because in uneven defects, the change in the amount of light between adjacent pixels is relatively gradual; This is because it is difficult to detect changes in the amount of light using processes such as differentiation. That is, by determining the average light amount for a plurality of pixels, changes in the light amount can be emphasized, and uneven defects can be easily detected based on changes in the light amount. Next, for the average light amount of one area, as shown in Figure 5, eight neighbors are determined, for example, for K ₅₃ , K ₄₂ , K ₄₃ , K ₄₄ , K ₅₂ , K ₅₄ , K ₆₂ , K ₆₃ , K ₆₄ The average light intensity of the 8 neighborhoods is calculated, and the difference in light intensity from K _MN is calculated among the 8 neighborhoods that are smaller than the light intensity of K _MN , and the largest difference is selected. Let this be _SMN . Expressed in the formula, S _MN = K _MN − MIN (8 neighbors of K _MN ) However, K _MN − MIN (8 neighbors of K _MN ) > 0,
MIN (8 neighbors of K _MN ) selects the minimum value among the 8 neighbors.

This S _MN is also calculated from S ₁₁ to S ₈₈ . However, S ₁₁
is calculated from the three neighborhoods of K ₂₁ , K ₂₂ , and K ₁₂ , and S ₁₂ is 5
Calculate from the neighborhood. S _N1 , S _N8 , S _1M , and S _8M are calculated in the same way. If MIN (8 neighbors of K _MN ) does not satisfy the condition, S _MN =0. By comparing the 64 S _MNs calculated in this way with preset criteria, it is possible to detect areas where uneven defects exist. Furthermore, when determining the quality of the inspected object 2 based on the presence or absence of uneven defects, rather than detecting the position of the uneven defects,
The purpose can be achieved by finding the maximum value of S _MN and comparing it with preset criteria.

To explain specifically, as shown in Figure 6a, K ₂₂
If there is a convex defect over K ₃₂ , the bright region
Assume that the average light amount of K ₂₂ is 100, and the average light amount of dark area K is 20. Further, if the remaining average light quantities are all 50, then S ₂₂ =80 and S ₃₂ =0 as shown in FIG. 6b, and a convex defect can be detected.

〔Effect of the invention〕

As described above, the present invention includes a light source that provides specularly reflected light to the surface of an object to be inspected, a television camera that captures a specularly reflected image of the object, and an A/D converter that converts the output of the television camera. D converter and the A/D
a frame memory that sequentially stores the output of the converter, and divides the image data stored in the frame memory into N pieces in the horizontal direction and M pieces in the vertical direction.
means for dividing an image into M×N regions; means for reading the image data of the one region and calculating an average value of brightness at each point; and means for comparing the average values of each region and calculating a difference; Uneven defects are detected using a comparison means that compares the difference between the calculated average values with a reference value, so even if the defects are not uniform in size, they can be processed using the same criteria, making them difficult to detect using conventional binarization processing. This has the effect of being able to detect uneven defects.

[Brief explanation of drawings]

FIG. 1 is a perspective view of an embodiment of the apparatus used in the detection method of the present invention, FIG. 2 is a front view of an example of a specularly reflected image of the television camera shown above, FIG. Figures a and b are explanatory diagrams of image data written in the frame memory, Fig. 5 is a diagram of the average light quantity allocation same as above, and Figures 6 a and b are explanatory diagrams of the unevenness defect detection method of the present invention. DESCRIPTION OF SYMBOLS 1... Light source, 2... Test object, 3... Television camera, 6... A/D converter, 7... Frame memory, 8... Computer, 9... Memory.

Claims (1)

[Claims] 1. A light source that provides specularly reflected light on the surface of the object to be inspected;
A television camera that captures a regular reflection image of the object to be inspected, an A/D converter that converts the output of the television camera from analog to digital, and a frame memory that sequentially stores the output of the A/D converter. , the image data stored in the frame memory is horizontally N
means for dividing one image into M×N regions by vertically dividing the image into M regions; means for reading out the image data of the one region and calculating the average value of the brightness of each point; and the average value of each region. 1. A method for detecting unevenness defects, characterized in that an unevenness defect is detected by a means for calculating a difference by comparing the average values, and a comparison means for comparing the difference between the calculated average values with a reference value.