JP3000615B2 - Pattern recognition method - Google Patents

Description

Description: TECHNICAL FIELD The present invention relates to a pattern recognition method, and more particularly, to a means for comparing a reference pattern having an unclear area with an inspection pattern.

(Prior Art) For example, pattern recognition for detecting the position and shape of an inspection object such as an electronic component mounted on a substrate or a position reference mark formed on the substrate is performed by using an electronic component or a position reference mark. This is performed by comparing a reference pattern obtained by observing an actual sample with a camera in advance with an inspection pattern obtained by observing an inspection target with a camera.

(Problems to be Solved by the Invention) By the way, when the sample or the inspection object is observed by an optical means such as a camera as described above, the light emitted from the light source partially mirrors the surface of the sample or the inspection object. In many cases, the entire shape cannot be clearly observed due to reflection or the like, and a partially unclear area often occurs, and the recognition accuracy tends to decrease due to the unclear area.

Therefore, an object of the present invention is to provide a pattern recognition method that solves the above-described problem.

SUMMARY OF THE INVENTION To this end, the present invention provides a method for recognizing a pattern including an optically unclear area due to at least partial specular reflection of the surface of a sample and an inspection object, comprising: Is observed by an optical means to detect an unclear area of the reference pattern, and then the inspection object is observed by the same optical means to detect the inspection pattern. When comparing the reference pattern with the inspection pattern, the unclear area is detected. Is excluded.

A method for recognizing a pattern including an optically unclear area caused by at least partial specular reflection of a surface of a sample and an inspection object, wherein the sample is observed by an optical means, and a sharp area of a reference pattern is observed. Then, the inspection object is observed by the same optical means to detect the inspection pattern, and when the reference pattern and the inspection pattern are compared, the clear area is compared.

(Operation) According to the above configuration, the unclear area is excluded, and the pattern matching is performed on the clearly observed area.
The recognition accuracy can be improved.

Example Next, an example of the present invention will be described with reference to the drawings.

FIG. 3 shows a state in which a character A formed on a sample 12 such as a substrate placed on a table 13 is observed by optical means such as a camera 10 and a ring light source 11. In FIG. 1, 1 is a sample taken by the camera 10 described above.
This is a reference pattern obtained by observing 12, and 8 × 8 = 64.
Pixel G. In this embodiment, the pattern of the character A is collated. However, the light emitted from the light source 11 is partially specularly reflected on the surface of the sample 12, and the entire shape of the character A is clearly observed. However, in the figure, the shaded portions are areas E11 and E12 that are obscurely observed and detected.

The inspection pattern 2 is a pattern obtained by placing an inspection target on the table 13 and observing it. Since this inspection object is also observed by the same optical means as that of the sample 12, areas E21 and E22 that are obscurely observed are formed in substantially the same portion as the reference pattern 1.

Therefore, the present means sets invalid pattern 3. The invalid pattern 3 corresponds to the pixels G corresponding to the areas E11 and E12.
The invalid areas E31 and E32 are set in the invalid areas E31 and E32.
Is excluded from the calculation described later.

Reference numeral 4 denotes a logic circuit (Exclusive or), and the reference pattern 1 and the test pattern 2 are connected to an input section of the logic circuit 4. Numeral 5 denotes a NAND circuit whose input section has a logic circuit 4 and a logic circuit 6 on the invalid pattern 3 side.
The NAND circuit 5 is connected to an operation unit 7. One input section of the logic circuit 6 is connected to the changeover switch section 8. The switch section 8 has a variable contact 81 and two fixed contacts 82 and 83 of 0 and 1. Each pixel G of each of the patterns 1, 2, and 3 has binarized or multi-valued data, and outputs the data of each pixel G to the logic circuits 4 and 6.

This means has the above configuration, and then the unclear area E11,
A pattern recognition method excluding E12 will be described. In this case, the variable contact 81 is connected to the fixed contact 82 on the 0 side.

Now, in the reference pattern 1 and the inspection pattern 2,
The binarized value of the pixel G into which the character A is taken is 1 and the other portions are 0. Therefore, when the 0,1 signals of the 64 pixels G of the reference pattern 1 and the inspection pattern 2 are sequentially input to the logic circuit 4, the data of both pixels G match at 0,0 or 1,1. The output of the logic circuit 4 is 0, and the output is 1 when there is a mismatch between 1,0 or 0,1. As described above, when the data of the pixels G of the reference pattern 1 and the test pattern 2 do not match and the output of the logic circuit 4 is 1, the output of the NAND circuit 5 is always 0. When the data of the two pixels G match and the output of the logic circuit 4 is 0, the output of the NAND circuit 5 is determined by the output of the other logic circuit 6. That is, in the invalid pattern 3, the invalid areas E31 and E3
The data of the pixel G other than 2 is 0. In this case, the input of the logic circuit 6 is 0, 0 and the output is 0.
The output of the ND circuit 5 is 1 and is input to the arithmetic unit 7.
However, since the pixels G of the invalid areas E31 and E32 are 1, the input portion of the logic circuit 6 is 1, 0, and the output is 1
Therefore, the output of the NAND circuit 5 becomes 0. This means that in the invalid areas E31 and E32, even if the data of the pixels G of the reference pattern 1 and the test pattern 2 match, the output of the NAND circuit 5 becomes 0 and the data is excluded from the calculation of the calculation unit 7. Means to be done. As described above, in the present method, the unclear invalid areas E31 and E32 are excluded from the calculation of the matching, so that the recognition accuracy can be improved. Note that the recognition accuracy is (number of matching pixels / invalid area
(Effective pixel number excluding the number of pixels of E31 and E32).

By the way, when the sample or the inspection object is observed by the optical means as described above, most of the observation is indistinct, and only a part of the sample or the inspection object is clearly observed. A matching method in the case will be described. In this case, the variable contact 81 is connected to the fixed contact 83 on one side.

In the reference pattern 1 of FIG. 2, the area E41 is clearly observed, but the other portions are obscurely observed. In the inspection pattern 2, this area
Only the area E42 almost similar to E41 is clearly observed. Therefore, an effective area E43 of the effective pattern 3 'is set for the area corresponding to the area E41 which is clearly detected, and matching is performed only on the effective area E43 as described below.

The input of the logic circuit 6 on the side of the switch section 8 is always 1, and the data of the pixel G other than the effective area E43 is 0 in the effective pattern 3 ', so that the output of the logic circuit 6 is 1. The output of the NAND circuit 5 becomes 0. This means that parts other than the effective area E43 are excluded from the calculation. In the effective area E43, when the data of the pixel G of the reference pattern 1 and the data of the pixel G of the inspection pattern 2 match, the output of the NAND circuit 5 is 1, and when the data does not match, the output is 0. It is. As described above, according to the present method, the matching calculation is performed only for the clear effective area E43.

(Effects of the Invention) As described above, the present invention excludes areas that are obscurely observed, and excludes only areas that are clearly observed.
Since the matching calculation is performed, the recognition accuracy can be greatly improved.

[Brief description of the drawings]

1 shows an embodiment of the present invention. FIG. 1 is a block diagram of a matching unit, FIG. 2 is a block diagram of another matching unit, and FIG. 3 is a perspective view of an observation device. 1 ... Reference pattern 2 ... Inspection pattern 10, 11 ... Optical means 12 ... Sample

Continuation of front page (58) Fields investigated (Int. Cl. ⁷ , DB name) G06T 1/00 G06T 7/00 G01N 21/88 H05K 13/08 H05K 3/00

Claims (2)

(57) [Claims] 1. A method for recognizing a pattern including an optically blurred area caused by at least partial specular reflection of the surface of a sample and an inspection object, wherein the sample is observed by an optical means and the reference pattern is blurred. In this case, the inspection area is detected by the same optical means, the inspection pattern is detected, and the inspection pattern is detected.When the reference pattern and the inspection pattern are compared, the above-mentioned unclear area is excluded and the comparison is performed. Characteristic pattern recognition method. 2. A method for recognizing a pattern including an optically unclear area caused by at least partial specular reflection of the surface of a sample and an inspection object, wherein the sample is observed by an optical means, and Detecting a clear area, then observing the test object by the same optical means to detect a test pattern, and comparing the reference pattern with the test pattern, the clear area is compared. Pattern recognition method.