JPH052647A - Pattern inspection method - Google Patents

Abstract

PURPOSE:To provide a pattern inspection method which automatically inspects the patterns of a graphic and a character, which are printed on a printed board, etc., with a highly reliable and easy method. CONSTITUTION:Picture data on a reference pattern is differentiated by individual directions. A differentiation result by the individual directions is binarized with a first low threshold and the binarized result is labelled. The differentiation result by the individual directions is binarized with a second high threshold. When the respective labels of a labelling result are in an effective part binarized by the second threshold, the label is decided and the number of labelling is calculated. The same signal processing is executed as to a pattern being the object of inspection. The obtained number of labelling is compared with that as to the reference pattern. When they coincide, it is set to be 'good'. When they do not coincide, it is set to be 'bad' so as to judge the 'good/bad' of the pattern.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a pattern inspection method for inspecting a graphic pattern by performing image processing, and more particularly to a pattern inspection method capable of automatically inspecting a graphic pattern by a simple method in a short time.

[0002]

2. Description of the Related Art There is a demand for a method capable of automatically inspecting defects of a wiring pattern on a printed circuit board, inspecting a CD printed surface, etc. in real time with high reliability. Such automatic inspection uses image processing method,
For example, a wiring pattern is input as image data, image processing is performed on the figure, and the quality of the wiring pattern is automatically determined based on the result. As a premise of such inspection, there is a demand for a highly reliable image processing method which has a simple method for real-time processing. If a highly reliable processing result is obtained by the image processing method, the reliability of the inspection using the result also becomes high.

As a conventional image processing method, a figure is input as two-dimensional image data by an image pickup device, and these image data are simply binarized by using one threshold value, and image density (intensity) information is obtained. There is known a method of performing binarized image processing with "1" or "0". There is also known an image processing method in which a direction-specific boundary is emphasized by applying a direction-specific differential and then binarized. The quality of the pattern is judged by comparing the results obtained by these image processing methods with the reference data.

Further, a method using grayscale image data is also known. As one of the methods of using the grayscale image data, there is a method of binarizing the differential result of the image data of the inspection object and comparing it with the binarized result of the preprocessed image data of the reference graphic. Are known. Further, as another method using the grayscale image data, a matching method based on correlation calculation is known.

[0005]

There is a method of using the area, the center of gravity, the perimeter, etc. as the feature quantity in the above-mentioned binarized image processing method, but the above-mentioned area and perimeter are several depending on the variation of the illumination condition. It fluctuates by 10%, and small defects cannot be detected. Further, it cannot be a feature amount for a defect whose center of gravity is small. Therefore, even if such an image processing method is applied, a defective pattern with high accuracy and high reliability cannot be detected.
In the image processing method based on the above-described direction-dependent differentiation, since the differential mask (matrix) is discrete (discrete) with respect to the direction, for example, when a mask with a step of 45 degrees is used, the intermediate value of 22.5 degrees is obtained. Since the boundary information has an intermediate gradation, the boundary may or may not appear at one specific threshold value, which may cause noise in the differentiation, resulting in a lack of reliability in image processing. The method of differentiating the grayscale image data and binarizing it cannot perform an accurate inspection if the reference pattern and the position of the inspection target are deviated. further,
Since the method of performing the correlation calculation is complicated in processing and the correlation coefficient does not change for small defects, it cannot be applied to detect pattern defects. Therefore, it is an object of the present invention to solve the above problems and provide a highly reliable pattern inspection method having a real-time processing property with a relatively simple method.

[0006]

In order to solve the above problems, a pattern inspection method according to the present invention differentiates two-dimensional image data of a pattern such as a figure or a character to be inspected from a plurality of directions according to directions. The direction-differentiated result is binarized with a first threshold value, the second-value binarized result is labeled, and the direction-differentiated result is a second threshold larger than the first threshold value. The effective pixel is binarized by the threshold value, and when the label specified by the labeling exists at the position of the effective pixel, the label is determined to be effective, and the determined number of labels is the reference number of labels. When they match with each other, it is determined that the pattern to be inspected has no defect.

[0006]

Function: The image data is differentiated according to direction, and the boundary according to direction is first emphasized. The direction-differentiated result is binarized with the first low threshold value, and the binarized result is labeled. In parallel to this process, the direction-specific differentiation result is binarized with a second threshold value higher than the first threshold value to determine an effective pixel. If each label based on the labeling exists at the position of the effective pixel, the label is determined to be effective. This number of labels is compared with the number of labels obtained as a result of the same processing as described above for the reference pattern, and if they match, it is determined that there is no defect. In this way, a highly reliable image processing result can be obtained by labeling the result obtained by binarizing the direction-specific differential with a low threshold value and matching it with the effective pixel binarized with a high threshold value. Then, the presence or absence of a pattern defect is determined by comparing the number of labels thus obtained with the number of labels for the reference pattern.

[0007]

FIG. 1 shows a pattern inspection apparatus for carrying out the pattern inspection method of the present invention. This image processing apparatus comprises an image inputting means 1, a direction-dependent differentiating means 2, a first binarizing processing means 3, a labeling means 4, a second binarizing processing means 5, a labeling confirming means 6, and a judging means. It consists of 7. The image input means 1 is realized by a CCD image pickup device or the like, and a pattern such as a figure or a character to be image-processed is, for example, 8
It is input as two-dimensional image data having a bit gradation. This pattern inspection apparatus first calculates the number of labels as a result of image processing on a pattern serving as a reference of a pattern to be inspected under favorable conditions. Next, the pattern inspection apparatus performs the same signal processing as the reference pattern on the inspection target, compares the number of labels with the number of labels of the reference pattern, and determines the presence or absence of a defect. Directional differentiation means 2, first binarization processing means 3, labeling means 4, second binarization processing means 5, labeling confirmation means 6,
The determining means 7 is realized by using a microcomputer in this embodiment. These processing functions will be described with reference to the flowchart of FIG. 2 and the graphic examples shown in FIGS.

First, the image processing for a pattern serving as a reference of the pattern to be inspected and the processing for calculating the number of labels as a result thereof will be described. Step S01: Image data input The image input means 1 inputs a figure to be image-processed as, for example, image data arranged in a two-dimensional manner with each pixel having an 8-bit gradation, and an image memory (not shown). )). FIG. 3 shows the result of displaying the image data stored in the image memory on the display device (not shown) in this way. In this figure example, a panel with a black background and a white figure printed on it is illuminated from the front of the panel, and the reflected light is input by a CCD image pickup device. In the white figure, each pixel has an 8-bit gradation ("2
55 ”), but there are few that actually show the value of“ 255 ”, and there is a large variation in the range of 0 to 255. Step S02: Differentiating by Direction The differentiating means by direction 2 performs differentiating by direction on the image data stored in the image memory. This direction-dependent differentiation processing is performed on the image data shown in FIG. 3 in four directions orthogonal to each other as shown in FIG. 4, N, E, S, W, and 4 between them.
Differentiating processing is performed sequentially from a 5 degree direction and a total of 8 directions using a known differential matrix (mask). FIG. 4 shows the direction-dependent differentiation result from the N direction among the direction-dependent differentiation. From this result, it can be seen that the boundary of the figure is emphasized in the direction along the N direction.

Step S03: 2 at low first threshold
Binarization The first binarization processing means 3 binarizes the direction-specific differentiation result with a first threshold value. FIG. 5A shows the first
"50" for the maximum gradation of "255" as the threshold of
The result of binarization with is shown. Step S04: Labeling process The labeling means 4 performs a labeling process on the binarized result. In this labeling, the continuity of the pixels (pixels) determined to be “1” as being larger than “50” in the binarization result is checked, and the cluster of consecutive pixels is set to 1
One label. FIG. 5A shows that the result of binarization with the first threshold value = 50 is labeled with 12 labels: label L1 to label L12. Step S05: Binarization with a High Threshold The second binarization processing means 5 is the first binarization processing means 3 described above.
In parallel with the processing with the labeling means 4, binarization processing is performed on the direction-specific differentiation result at a second threshold value = 70 higher than the first threshold value to determine an effective pixel. FIG. 5B shows the result of binarization with the second threshold value. When binarization is performed with a high threshold value, there will be a portion where the portion that is set to “1” with a low threshold value disappears. In this example, a noise component appears in the direction-dependent differentiation from the N direction.
The portion corresponding to the labels L9 to L12 in (A) is set to "0" at the second threshold value.

Steps S06 to S09: Labeling Confirmation Processing The labeling confirmation means 6 uses the label L shown in FIG.
For 1 to the label L12, it is determined whether or not the portion where the label is located is the portion of the effective pixel which is "1" in the binarization processing by the second threshold value by the second binarization processing means 5. Yes (step S06). If the label has an effective pixel portion which is "1" as a result of binarization with a high threshold value, the label is determined to be correct (step S07). On the other hand, if the label is a portion which is "0" as a result of binarization with a high threshold value, the label is erased because it has a high noise component (step S08). Labeling confirmation processing is performed for all labels L
It performs about 1-L12 (step S09). In this example, the labels L1 to LL8 are confirmed and the labels L9 to L12 are
Was erased. Step S10: Label organizing The labeling organizing means 6 arranges the labels by advancing and labeling the next numbered label when there is a label which is erased in the middle of the label after the labeling is established.

Binarization with the first threshold value for the above-mentioned direction-dependent differentiation result, its labeling processing, determination of effective pixels by binarization with the second threshold value, and labeling confirmation processing are other Do the same for directions. With the above processing, the number of labels in each direction can be obtained for the reference pattern.

Next, with respect to the pattern to be inspected which is actually inspected, the same processing as in steps S01 to S10 is performed. As an example, as shown in FIG. 6, a defective graphic pattern having protrusions at the A and B portions with respect to the reference pattern shown in FIG. 3 will be described as an example. FIG. 7 shows the direction-dependent differentiation result from the N direction. Defects A and B are also emphasized by this direction-wise differentiation. The direction-based differential results of the defective portions A and B are not continuous. The result of binarization of the differentiation result by direction with the first threshold value = 50, and the result of further labeling processing, and the result of binarization with the second threshold value = 70 to determine the effective pixel, and the result of confirming the labeling Figure 8
Shown in. The result of FIG. 8 is the same as that of FIG.
The portion of the definite label L4 shown in (C) is the label L41,
The label is divided into three labels L42 and L43, and the number of labels is increased to 10.

Steps S11 to S13: Judgment of the Number of Labels The judgment means 7 determines the number of labels obtained for the reference pattern =
8 and labels L1 to L obtained for the above inspection target
The number of labels of 8, L41, and L42 = 10 is compared and they do not match, so it is determined that the inspection target has a defect. This judgment is very easy because it is a comparison of the number of labels.
The image processing method described above is simple and can be processed in a short time.
Moreover, it is highly reliable.

In carrying out the present invention, various modifications other than those described above can be taken. For example, the above-described embodiment of the present invention has been described with respect to the case where the front side of the graphic pattern is illuminated and the reflected light thereof is input as image data. May be processed. Further, the present invention can be applied not only to figures but also to pattern inspection of characters and the like.

[0015]

As described above, according to the present invention,
A pattern inspection method to which highly reliable image processing is applied can be realized by a relatively simple method. Since the image processing method and inspection method of the present invention described above are based on a simple method,
It can be realized with a simple configuration, can obtain results in a short time, and has real-time performance. Therefore, for example, it can be effectively applied to online pattern automatic inspection of printed circuit boards, CDs and the like.

[Brief description of drawings]

FIG. 1 is a configuration diagram of a pattern inspection apparatus that implements a pattern inspection method according to an embodiment of the present invention.

FIG. 2 is a processing flowchart of the pattern inspection apparatus shown in FIG.

FIG. 3 is an image display diagram of a reference pattern subjected to image processing by the pattern inspection apparatus of FIG.

FIG. 4 is a diagram showing a result of the direction-specific differentiation processing of the image data shown in FIG.

5 is a diagram showing a binarization result of the result shown in FIG. 4 in the pattern inspection apparatus of FIG.

6 is an image display diagram of a defective inspection target pattern subjected to image processing by the pattern inspection apparatus of FIG.

FIG. 7 is a diagram showing a result of the direction-dependent differentiation processing of the image data shown in FIG.

8 is a diagram showing a result of binarizing and labeling the result shown in FIG. 7 by the pattern inspection apparatus of FIG.

[Explanation of symbols]

1 ... Image input means, 2 ... Directional differentiation means, 3 ... 1st binarization processing means, 4 ... Labeling means, 5 ... 2nd binarization processing means, 6 ... Labeling confirmation Means, 7 ・
-Judgment means.

Claims (1)

Claim: What is claimed is: 1. A two-dimensional image data of a pattern such as a figure or a character to be inspected is differentiated from a plurality of directions according to directions, and a result of the differentiation according to the directions is a first threshold value. In 2
Quantize and label the binarized result,
When the effective pixel is determined by binarizing the direction-differentiated result with a second threshold value larger than the first threshold value, and the label defined by the labeling is present at the effective pixel position. A pattern inspection method for determining the label, comparing the determined number of labels with a reference number of labels, and determining that the pattern to be inspected has no defect when they match.