JPH01150987A - Method for recognizing shape - Google Patents

Abstract

PURPOSE:To accurately recognize the existence of a defect, etc., by preparing plural recognizing patterns with respective different patterns and recognizing the shape of a recognization object substance according to the synthetic result based on the matching of a picture with all these prepared plural recognizing patterns. CONSTITUTION:When a synthetic decision related to a defective part A is executed, in a CPU 4, it is not decided that 'a defect exists, and a defect type is A' only by the matching of one recognizing pattern 20a and a binarizing picture corresponding to the defective part A, but the synthetic decision is executed by synthesizing detected results of the matching of the other recognizing patterns 20b-e and the binarizing picture with that of the above-mentioned matching. In this case, in the CPU 4, when it is decided that 'no defect exists' with a recognizing pattern 20e, and moreover, it is decided that 'a defect exists, and a defect type is A' with at least three recognizing patterns or more out of the recognizing patterns 20a-d, this is made into a logic composed as 'a defect exists, and a defect type is A', and the synthetic decision is executed. Thus, the existence of the defect in the recognition object substance can be recognized correctly.

Description

DETAILED DESCRIPTION OF THE INVENTION [Field of Industrial Application] The present invention relates to a shape recognition method for recognizing the presence or absence and shape of defects such as wiring patterns on printed circuit boards.

[Conventional technology]

Conventionally, a reflected light image or a captured image of a predetermined resolution obtained by scanning the wiring pattern of a printed circuit board with a light beam or capturing an image with an imaging device is processed in advance according to the logic for detecting protrusions, breaks, etc. It is examined using a shape recognition pattern with a predetermined resolution, and the presence or absence of defects is determined, such as parts where the line width of the wiring pattern shown in the reflected light image or captured image is less than a tolerance value are chipped, and parts where the line width is more than the tolerance value are protrusions. There is a shape recognition method that recognizes.

[Problem that the invention seeks to solve]

However, since the conventional method described above uses a single recognition pattern, for example, the broken line portion in FIG. 5(a) may be recognized as a missing part of the wiring pattern, or
) is recognized as a right-angled section with no defects, and defects are often overlooked and over-detected, resulting in a lack of reliability.

An object of the present invention is to provide a highly reliable shape recognition method that can accurately recognize the presence or absence of defects.

Means for Solving Problem C] In the present invention, a plurality of recognition patterns each having different patterns are prepared, and according to the overall result based on matching the image with all of the plurality of prepared recognition patterns,
The shape of the object to be recognized is recognized.

[Effect]

This prevents defects from being overlooked or detected excessively, which would occur if only a single recognition pattern was used.
More accurate recognition becomes possible.

〔Example〕

FIG. 1 is a block diagram illustrating an embodiment of the method according to the invention. In the shape recognition apparatus shown in the figure, the scanning optical system 1 scans the wiring pattern surface of the printed circuit board PWB to be recognized with a laser beam and outputs an image signal of the reflected light image. A polygon scanner 10 that collects the reflected light image from the wiring pattern surface, a condensing optical fiber 11 that condenses the reflected light image from the wiring pattern surface, and a condensing optical fiber 11 that synchronizes the reflected light image condensed by the condensing optical fiber with the scanning speed of the polygon scanner 10. It is composed of an AD conversion module 12 that samples the image at a high speed and then outputs it as a binary image signal.

The image memory section 2 stores the binarized signal output from the scanning optical system 1 in units of pixels with a predetermined resolution.

Here, as an example, the image memory section 2 has 11 rows x 11 rows.
It is assumed that a storage capacity for columns, that is, 11×11 pixels is provided.

That is, when a binary signal is output from the optical system 1, the binary signal is sequentially stored in each shift register of the pipeline module 20.

The line buffer section 21 is composed of an 11-line line buffer, and sequentially sends the image information stored in each line of the pipeline module 20 to the next line of the same module 20.

That is, at the time when one scan is completed by the polygon scanner 10, a binarized signal corresponding to the pixel row corresponding to the minus scan is stored in the top stage of the pipeline module 20. Thereafter, each time one scan is completed by the scanner 10, the line buffer unit 21 stores the binarized signal stored in the i-th row of the pipeline module 20 in the i+1 row of the same module 20. .

Then, when the polygon scanner 10 completes 11 scans, the pipeline module 20 stores the binarized information for the 11 scans, and the pipeline module 20 is thus filled with the binarized information. After that, each time one scan of the scanner 10 is completed, a binarized signal corresponding to the pixel row for that scan is sent to the module 2.
0, and the contents stored in each row of the same module 20 are sequentially transferred and stored in the rows.

As a result, the pipeline module 20 stores board information of 11 x 11 pixels (assuming that the shaded area indicates the wiring pattern of the board) as shown as 20a to 20e in FIG. 1, and its contents. is updated sequentially as the above scanning is performed.

Next, the shape recognition section 3 shown in the same figure recognizes the shape of the wiring pattern based on the contents of the image memory section 2, and R
Consists of AM30A to 30E, with missing wiring patterns,
A signal indicating the presence or absence of defects such as thick lines and protrusions is output.

That is, defects in the wiring pattern can be recognized according to the combination of the 11 x 11 pixel binarized images, and each address corresponding to the combination of the 11 x 11 pixel binarized images can be recognized. The predetermined shape recognition logic is
If AM30A to AM30E are stored in table form in advance, the contents of the image memory section 2 can be stored in each RAM3.
By supplying the signals 0A to 30E as address inputs, these RAMs 30A to 30E output signals indicating the determination contents corresponding to the shape recognition logic contents for each shape of the wiring pattern.

However, in this case, if it is attempted to determine all combinations of 11×11 pixels, the number of addresses will be extremely large. Therefore, in this embodiment, recognition patterns a to e as shown by O marks in FIG. It is configured such that only the binary signals of the pixels are used as address inputs of the RAMs 30A to 30E, respectively.

For example, the RAM 30A performs shape recognition based on the recognition pattern a shown in FIG. 2(a).

That is, the cross-shaped recognition pattern a shown in FIG. 2(a)
When set, only the address information of the module 20 indicated by a circle at 20a in FIG. 1 is input to the RAM 30A.

On the other hand, the RAM 30A is preset with shape recognition logic that can determine the shape of the input image based on the combination of the addresses of the cross-shaped recognition pattern a.

Specifically, each length measuring element U shown in FIG. 2(a).

Shape recognition logic is stored corresponding to combination patterns of "1" and "0°" of D, R, L, and P.

For example, when a wiring pattern image 100 with a portion missing as shown in FIG. 3 is viewed using the cross-shaped recognition pattern a, the missing portion will be “1” and the other portions will be “0#”.

In this case, each length measuring element U, D, R, L, P is U-11
111, D-11000°R-11000, L-11000, P-1. Therefore, in RAM30A, these length measuring elements "1" and "0"
The shape recognition logic corresponding to the combination pattern outputs a signal indicating whether or not a given condition is satisfied.

The same applies to other recognition patterns, and RAM30B
~30E also have shape recognition logic set in advance that outputs recognition information according to each corresponding pattern.

The CPU 4 comprehensively determines the shape of the wiring pattern based on the outputs of the RAMs 30A to 30E.

That is, in the RAMs 30A to 30E, defect determination is performed by comparing the recognition patterns a - e with the binarized images, respectively, but in the CPU 4, the results of these comparisons are comprehensively judged and accurate defect determination is performed. It will be done.

FIG. 4 shows defective parts A, B, and C of the wiring pattern and normal parts of the same pattern.

The table below shows the results of comparing the defective parts A, B, and C of the above wiring pattern and the normal parts of the same pattern with the recognized patterns a to e. An x mark indicates a case where it is determined that there is no defect based on the recognition pattern, and an x mark indicates a case where it is determined that there is a defect based on the recognition pattern.

For example, regarding defective part A, recognition pattern a
It can be seen that, while the defect is determined correctly by -d, the defect is overlooked by the recognition pattern e.

Also, if you look at the normal part, ′, ``pattern a, b''
Although it is correctly determined that there is no defect by and -d, over-detection of defects may occur due to the recognition pattern c and 'el'. “In this way, it can be seen that the combinations of determination results for each recognition pattern differ depending on the type of each defective portion and normal portion of the wiring pattern.

Therefore, a defect detection algorithm is configured in the CPU 4 based on the above combinations.

For example, when comprehensive judgment regarding defective part A is performed, the CPU 4 uses one recognition pattern a and defective part A.
It is not determined that "there is a defect and the defect type is A" only by comparing the binarized images corresponding to , but a comprehensive determination is made in conjunction with the detection results of other recognition patterns b to e. In this case, the CPU 4 determines that there is no defect based on the recognition pattern e, and when it is determined that there is a defect and the type of defect is A based on at least three of the recognition patterns a to d. “Defective”
The logic is constructed such that the type of defect is A'', and a comprehensive judgment is made.

The comprehensive determination result obtained by the CPU 4 in this way, that is, a signal representing the presence or absence of a defect, the determined type of defect, and the coordinate position, is displayed on a display unit (for example, a CRT is used), not shown.

As described above, according to this embodiment, each of a plurality of different recognition patterns is compared with a binary image, and the shape of the wiring pattern is recognized according to the overall result based on these comparisons.
Note that the logical configuration of the CPU 4 may be such that the detection results that are the majority of the detection results based on various recognition patterns are determined to be the correct determination results.

In this case, when the CPU 4 determines that "there is a defect and the defect type is A" based on three or more recognition patterns, which is the majority of the various recognition patterns, the CPU 4 changes this to "there is a defect and the defect type is A", and so on. A comprehensive judgment will be made. Of course, as the algorithm for this determination in the CPU 4, an algorithm that comprehensively determines only the presence or absence of a defect may be adopted, regardless of the type of defect.

Further, in the embodiment, the shape of the wiring pattern of the printed circuit board is recognized, but the method according to the present invention is not limited to this, and can recognize any type of paper, plastic, glass, film, cloth, various metals, etc. It is applicable to shape recognition of a target object.

Further, the shape, number, etc. of the recognition patterns are also arbitrary, and the shape, number, etc. of these recognition patterns are freely selected depending on the shape of the object to be recognized, etc., respectively.

Furthermore, in the embodiment, images are obtained by scanning a laser beam, but the invention is not limited to this.
A CCD linear sensor or the like may be used, and any method for obtaining an image from the object to be recognized may be used.

〔Effect of the invention〕

As explained above, according to the present invention, a plurality of different recognition patterns are prepared, and the shape of the recognition target object is determined according to the overall result based on matching these recognition patterns with images of the recognition target object. can be recognized accurately.

Furthermore, when inspecting the object to be recognized for defects, overlooking or over-detecting defects can be effectively prevented, and reasonable and accurate defect detection can be achieved.

[Brief explanation of the drawing]

FIG. 1 is a block diagram showing an example of a device used to carry out the method according to the present invention, FIG. 2 is a schematic diagram illustrating a recognition pattern in the embodiment, and FIG. 3 is a block diagram showing an example of a recognition pattern in the embodiment. FIGS. 4 and 5, which are explanatory drawings for explaining the detection method, are partial plan views showing an example of a defect in a printed circuit board wiring pattern and an example of a pattern in which the defect is easily misidentified, respectively. 1... Scanning optical system, 2... Image memory unit, 3...
Shape recognition unit, 4...CPU. DESCRIPTION OF SYMBOLS 10... Polygon scanner, 12... AD conversion module, 20... Pipeline module, 30A-30E... RAM. Figure 2 0. 0 υ ℃

Claims (1)

[Claims] Shape recognition in which a recognition pattern for shape recognition is prepared for an image of a predetermined resolution obtained regarding an object to be recognized, and the shape of the object to be recognized is recognized based on matching this recognition pattern with the image. In the method, a plurality of the recognition patterns each having different patterns are prepared, and the shape of the object to be recognized is recognized according to a comprehensive result based on matching the image with all of the plurality of prepared recognition patterns. A shape recognition method characterized by: