JPH0814847B2 - Pattern inspection method - Google Patents

Description

The present invention relates to a pattern inspection method for automating visual appearance inspection of a printed circuit board.

[Prior Art] As seen in Japanese Patent Laid-Open No. 63-15380, a repeated pattern is formed on a binarized image or a grayscale image.
There is a method of measuring the address of the pattern at a preset position and inspecting the average position and the alignment of the repeated patterns.

[Problems to be Solved by the Invention] However, in the conventional method, since the measurement position has to be set in advance for each individual pattern, there is a problem that the setting work becomes complicated when a plurality of patterns exist. In addition, when a work set or a position shift due to the deformation of the pattern itself occurs, the position where the measurement is deviated from the original measurement position is measured, and a correct measurement result cannot be obtained.

The present invention has been made in view of the above problems, and its object is a pattern inspection method that does not require a complicated data setting work in advance and can perform a pattern inspection correctly regardless of the presence or absence of a positional deviation of an image. To provide.

[Means for Solving the Problem] According to the present invention, a grayscale image composed of a pattern of a repetitive shape is differentiated into a fine line to obtain an edge image, and a direction code of a density gradient in a pattern portion of interest in the edge image is set as an X axis. Each direction code value obtained by projecting in the direction or the Y-axis direction is stored in memory, and only the direction code value of interest is added with reference to the stored direction code value to obtain a histogram. is there.

[Action]

Thus, according to the present invention, a grayscale image having a pattern of a repetitive shape is differentiated into a thin line to obtain an edge image, and the direction code of the density gradient in the target pattern portion of the edge image is set in the X-axis direction or Y direction. Each direction code value that is obtained by projecting in the axial direction is stored in memory, and only the direction code value of interest is added to obtain the histogram by referring to this stored direction code value, so check from the statistical value. Therefore, complicated data setting such as a method of recognizing individual pattern shapes is not required, and since individual pattern shapes are not recognized, measurement error due to individual pattern displacement does not occur.

〔Example〕

 The present invention will be described below with reference to examples.

FIG. 1 shows the configuration of an embodiment of a pattern inspection apparatus using the method of the present invention. In this embodiment, a video signal obtained by picking up an image from above the printed circuit board 1 by the TV camera 2 is A /
It is converted into a digital signal through the D converter 3.

The digitally converted signal is differentiated by the differentiating circuit 4,
The absolute value of the derivative representing the concentration gradient and the direction code representing the direction of the concentration gradient are obtained. The obtained absolute value of the differential is binarized by the binarization processing circuit 5 so as to represent the shape of the pattern, and further thinned by the thinning processing circuit 6 to obtain edge image data. The edge image data obtained by the thinning processing circuit 6 is written in the frame memory 7.

 The direction code is also stored in the direction code memory 8.

The edge image described above is a good representation of the shape of the pattern as shown in FIG. Here, in FIG. 2A, the arrow at the tip of the circuit pattern P indicates the direction of the direction code and indicates a part of the pattern shape.

Next, a method of creating the direction code histogram in the determination processing unit 9 will be described with reference to FIGS.

First, the portion of the pattern of interest on the obtained image is determined. For example, when it is desired to know the displacement of the image in the X-axis direction, attention is paid to the tip of the circuit pattern P.

Next, the direction code of the portion of interest is examined, and only the direction code of the portion of interest is projected in the axial direction that well represents the desired information. In FIG. 2A, if it is desired to know the positional deviation in the X-axis direction, the direction code at the tip of the circuit pattern P is projected in the X-axis direction. The code histogram obtained by this projection is shown in FIG.

FIG. 4 shows the flow of this projection processing. The edge image on the frame memory 7 is scanned, and when the edge image is detected, the direction code value on the direction code memory 8 corresponding to the address on the edge image is referred to, and if it is the direction code of interest, it is added to the histogram.

When the histogram of the direction code is obtained, statistics such as the median of this histogram and the width of the distribution are calculated. The circuit pattern P is inspected based on these statistics. That is, the histogram statistic of the direction code of the non-defective circuit pattern is registered in advance, and is compared with the statistic of the histogram obtained by the measurement as described above to obtain the circuit pattern P.
Can be inspected.

For example, from the median value, as shown in FIG.
The position of the circuit pattern P in the X-axis direction is obtained. From the distribution width, for example, if the distribution width is as shown in FIG. 3 (b), the tip of the circuit pattern P is as shown in FIG. 3 (a). It can be seen that are not aligned. That is, the degree of alignment of the tips of the circuit patterns P can be inspected by the width of the distribution.

〔The invention's effect〕

According to the present invention, a grayscale image having a pattern of a repetitive shape is differentiated into a thin line to obtain an edge image, and the direction code of the density gradient in the pattern portion of interest in the edge image is set in the X-axis direction or the Y-axis direction. Each direction code value obtained by projecting is memorized, and only the direction code value of interest is added to obtain the histogram by referring to this memorized direction code value. No need for complicated data setting such as the method of recognizing individual pattern shape, and because individual pattern shapes are not recognized, measurement errors due to individual pattern deviations do not occur and image misregistration on the right It has an effect that the pattern can be accurately measured regardless of the right side.

[Brief description of drawings]

FIG. 1 is a system configuration diagram of an embodiment using the method of the present invention,
2 (a) is an example image of the edge image obtained by the same as above, FIG. 2 (b) is an example diagram of the direction code histogram of FIG. 2 (a) obtained by the same as above, and FIG. 3 (a) is the same as above. FIG. 3 (b) is an example diagram of the direction code histogram of FIG. 3 (a) obtained by the above, and FIG. 4 is a flowchart of the projection processing of the same. 1 is a printed circuit board, 2 is a TV camera, 3 is an A / D converter, 4 is a differentiation circuit, 5 is a binarization processing circuit, 6 is a thinning processing circuit,
7 is a frame memory, 8 is a direction code memory, 9 is a determination processing unit, and P is a circuit pattern.

Claims (1)

[Claims] 1. A grayscale image having a repeating pattern is differentiated into a thin line to obtain an edge image, and a direction code of a density gradient in a pattern portion of the edge image to be projected is projected in the X-axis direction or the Y-axis direction. The pattern inspection method is characterized in that the obtained direction code values are stored in memory, and the stored direction code values are referenced to add only the direction code values of interest to obtain a histogram.