JPS58134372A - Pattern checking device - Google Patents

Abstract

PURPOSE:To attain the mass productivity, by securing free setting to a storing circuit with a keyboard input for the conditions of shape, position, number, etc. of window regions, various threshold values, correlative conditions, various deciding conditions, etc. of each subject pattern. CONSTITUTION:A feature extracting circuit 53 receives a dot signal from a dividing circuit 52 and extracts the feature of the signal as well as counts the number of signals by a counting circuit 54 to obtain an area. The counting circuits 54 and 54' perform counting under the control of a window region signal, and the numbers of window regions feature items, etc. are rearranged to be stored temporarily in a storing circuit 55. A window region generating circuit 63 produces signals showing many window regions of optional shapes which are edited on the basis of the data fed to a setting value storing circuit 65 from a keyboard 66. The feature data of each window region which is read out of the circuit 55 is compared with the upper and lower limit set values alphaij and betaij which are set for each window region and feature through a primary deciding circuit 56. Then a primary decision is given to check whether the feature data is within upper and lower limit ranges.

Description

DETAILED DESCRIPTION OF THE INVENTION The present invention relates to a pattern inspection device that determines the quality of a pattern to be inspected.

1st Cabinet《Sister. @ is 1, which is the principle of a conventional pattern inspection device called a template matching method set! In the Ming Pavilion, 0 Okasho, l and 1' are f, and inspection target patterns 12 to 8 are sampling points indicated by 10 marks, respectively.

Assume that in the first fW(a), the pattern l to be inspected is the letter 8. At this time, using a template (not shown), sampling points 2 to 8 are determined as shown, and the presence or absence of a pattern in each India is inspected. For example, if there is a pattern, it will be 1%, and if there is no pattern, then it will be rongj.
The inspection results at each point 2 to 8 are coded as AO, and from the code it is recognized which known pattern the pattern to be inspected is classified into.

Figure 1 (-) The inspection result is positive (logic 1) for any of points 2 to 8, and in this case, the pattern l to be inspected is □ character 8. Figure 1 et al.)
In , only point 8 has no pattern (logical 0), and in this case, the Arboris person for determination is determined in advance as a pattern of the character 0, and O is recognized accordingly.

In this pattern inspection method using template matching, the information to be examined is compressed and small (only the presence or absence of a pattern at a sampling point);
It has the rumored advantage that it is easy to classify patterns based on the results of examination, and it is possible to easily classify unknown pattern patterns to be inspected, such as with OCR (optical character reader).
(Known) It is known that it is effective when used in a pattern 1IIIli device where it is sufficient to determine which of the patterns corresponds to the pattern. When applied to If the judgment is made based on whether the pattern exists or not, there is a fatal flaw in that a similar but defective pattern is mistakenly judged as a good product. In the formula, if the pattern does not exist in the unique area, it is determined to be defective, but
This is because if a pattern exists but its size indicates that it is a good product, it cannot be determined that it is defective.

In commonly printed patterns, the surface area and area of the pattern portion in the ygA area vary greatly depending on the location, and each pattern has its own characteristics (in this case, 1 space 11 # □) in each window area. □), therefore, correct pattern inspection cannot be performed unless this feature is extracted and measured.

The present invention has been made in view of the conventional technical circumstances as described above, and therefore, the purpose of the present invention is to develop it based on the original template matching method and to apply it to general printing. It is an object of the present invention to provide a pattern inspection device that can correctly inspect patterns.

The outline of the present invention is as follows. In the present invention, first, a pattern to be inspected is scanned in two-dimensional sequential manner by a charging conversion device (
For example, the time-series electrical signals □ obtained by capturing an image using an industrial television camera are processed to determine the quality of the pattern. A region is set by electrical means or the like, and the features of the pattern portion viewed from each window region are quantitatively extracted.

On the other hand, set the upper and lower threshold ranges for each chapter area, and if the extracted waiting amount is within that range, it will be 111%, then -1 [6] −
Some of the 0-order window areas are grouped into groups. 1 For each group, a set of primary judgment results for each of the outer regions belonging to 1 is obtained based on the preset order N! 1 table (
This is done by dividing the inspection cost into window areas as in 1 for patterns that are non-defective products, obtaining the group-by-group gathering of the primary judgment results obtained for each area in advance, and making a table for comparison). 1 Based on this, a judgment is made as to whether the pattern is true or false (this is called a secondary judgment) - This secondary judgment is carried out for all groups, and based on the results, the truth or falsehood of the pattern is judged comprehensively - In addition, in the above secondary judgment, For groups determined to be defective, correlate ml as primary 1l111 as necessary! Do the following. The sse judgment is as follows.

Even if the inspection target pattern is a bird product, if it deviates from the predetermined position within the camera field of view, the feature values extracted through the predetermined window area may not fall within the predetermined threshold range. \As a result, - the soft decision result is bad. 0 However, since the 1 pattern is only slightly shifted to the adjacent one, the feature quantity detected through the predetermined window area and 1 the feature quantity detected through the proud window area. If the correlation is taken (in this case, correlation means addition, subtraction, etc.), the obtained correlation amount will fall within a predetermined threshold range, and the first judgment result will be good. Based on this idea, we determine an appropriate combination of related window areas, use the combination to perform arithmetic processing (addition if kept) of each feature, and perform a first-order judgment on the processing results (this is called a correlation-order judgment). say). By comparing and collating the set of correlation/soft decision results with a preset correlation/secondary decision table (predetermined in the same manner as the previous first order decision table), a second correlation decision is made. Judgment is made based on the results.

As explained below, in the present invention, - next determination,
In addition to the secondary judgment, the quality of the pattern to be inspected is determined by performing correlation judgment as necessary. In addition,
The shape, size, etc. of the window area are determined arbitrarily and appropriately according to the pattern to be inspected, and are not limited to a rectangular shape or the like. Moreover, as the feature quantity extracted through the window area, in addition to the area of the brim turn part in the room area Jd, the circumference 1 of the pattern part and other scales are of course used.

Next, an embodiment of the present invention will be described with reference to the drawings.
(a) is an explanatory diagram showing an example of setting the window area in the first invention, FIG. FIG.

In these figures, 21 to 27 are short 1!1 window areas, and ho is the inspection layer pattern (letter 8).
). As seen in 1112g(c), a portion of one pattern '□''-20 is seen through each of the window areas 21-27.

The third m is an explanatory diagram showing another setting example of the window area◇qm
In -jlZ, window areas 31 to 36 are set outside the same window areas 21 to 27 as shown in diagram (1), and window areas 41 to 48 are set inside, If the pattern placement is shifted, etc., the central window area (21
+27) and the outer or inner so area.

FIG. 4 is a block diagram showing an embodiment of the present invention. In the same figure, 50 is a pattern to be inspected, 51 is an industrial television camera (ITV camera), 52 is a two-part pixel division circuit, and 53153' is a pixel division circuit. That's the feature extraction circuit, 54e5
4' is fre counting circuit 1! s3 is a count value storage circuit, 5
6 is a primary judgment circuit, 57 is a secondary judgment circuit, 58 is a correlation judgment necessity circuit, 5.9 is a correlation calculation circuit, 60 is a correlation-order judgment circuit, 61- is a secondary judgment circuit between phases, 62 is a comprehensive judgment 63 is a we area generation circuit, 64 is a control circuit, 65 is a set value storage circuit, and 66 is a keyboard 1.

Next, an overview of the operation will be explained. First, the target pattern 50 is
The time-series electrical signals obtained by imaging with the TV power source 51 are binarized in the binarization/pixel division circuit 52 and divided into pixels (dots). The first feature extraction circuit 53 receives the dot signal from the one division circuit 52, which divides it into 320 dots in the horizontal X-axis direction and 240 dots in the vertical Y-axis direction, about 77,000 dots in total.
Extract that opportunity. For example, if the feature is simply the size of the area, focus on the dot signal's white if it is white, or its black if it is black. The area is determined by counting in the counting circuit s4.

(Find the area by meeting the meeting lengths). 53' is a circuit that extracts other features other than area, for example, when detecting a white and black boundary layer, it is a circuit that extracts the lump rising point, and the number of mlF points is calculated by the first-order counting circuit 54'. The JII field length is found by counting. Since the counting circuits 54 and 54' perform the St- operation under the control of the window area signal generated from the window area generating circuit 63, the needle interval result is as follows. This is the feature quantity extracted for each window area, which is arranged in the window area number and timing 1Iili111 and temporarily stored in the storage circuit sS. The room area generation circuit 63 can generate and output signals representing a large number of window areas of arbitrary shapes edited based on data input from the keyboard 66 to the set value storage circuit 65.

Now, when the scanning of one field by the ITV camera 51 is completed, the measurement and storage of the characteristic data for each window area are completed, and based on these characteristic data, the following circuits 56 to 65 determine whether the target pattern is good or not. Determine.

The feature data for each window region successively outputted from the memory circuit 55 is processed by the soft decision circuit 56 into an upper limit setting value αij1 and a lower limit setting value βij (however,
i represents the type of feature and j represents the number of the window area), and I5 and I5 within the upper and lower ranges (threshold range)
A primary determination is made as to whether or not there is. The determination processing operation for feature item ii will be specifically explained. ′
゛-Linear constant 1 path tI6 is the data D in the window area j regarding the amount of feature items (for example, the size of the rkJ product in the case of area)
A primary judgment is made at the separator to determine whether or not ij is within the range of the upper limit set value βij, and this is performed for the entire window area.〇-Primary judgment result, range of upper and lower thresholds If it is within the range, it is expressed as 6mi, and if it is outside the range, it is expressed as logical 0.The upper and lower thresholds can be selected appropriately for each window area and each of the 41 items. Since the eclipse data is entered manually from the keyboard 66 and stored in the set value memory path 65, the soft decision circuit S
6 reads and uses this 0 In general, it is often possible to think that an inspection layer pattern is composed of N small patterns (N is any integer), so each window area consists of N small patterns. The images are grouped into N groups in proportion to the number of layers.

For example, if the pattern to be inspected is ``A B CDJ'', it is assumed that it consists of 4□'- small patterns (each character makes up one small pattern):
, Ru. Note that "small pattern" does not necessarily mean 1wJ in units of 1 character.
It is not composed of a shape unit, but is sometimes composed of multiple character units.

The combination of grouped electric range symbols is on the keyboard 66
The set value is manually inputted using the set value storage circuit 65 and stored in the set value storage circuit 65. The secondary determination circuit 57 successively outputs the grouped window-area number combinations from the storage circuit 65, and divides the primary determination results for each window area from the primary determination circuit 56 into groups accordingly. Next, the secondary judgment (b) path 57 compares and checks the grouped primary judgment circuits 56 with the primary judgment table read out from the electric setting value storage circuit 65 to judge the quality of one group. Note that the next judgment table is
This is determined in advance for each target pattern and stored in the set value storage circuit 65.

The primary determination table when the window area set as shown in FIG. 13 is used and the numbers 8 (FIG. 2 6) and 7 are used as target patterns is shown in FIG. s.

When the target pattern is the number 8, the -th order determination table is coded data C1 arranged in the order of the window area symbols.
111111"...0], and when the number 7 is (1110000...0), it will be understood from FIG.

The secondary determination operation by the secondary determination circuit 57 is performed for all groups. Poor secondary judgment as a group (N
G), the correlation determination necessity circuit s8 determines whether correlation determination is necessary or not. The correlation determination necessity circuit s8 determines whether correlation determination is necessary or not according to the preset Ik conditions. If it is determined and becomes 1, it does not reach that (
In other words, when it becomes necessary to determine the correlation,
Correlation calculation 1iinss ~ combination of set window areas (
Data regarding this combination can also be obtained by reading out from the memory circuit 6B), and the calculation of the extracted feature data (
For example, the data in the window area 21 and the data in the window area 31 are added, and the result is output as new data DC1y.

The new data D Cjk obtained by the calculation is sent to the correlation-soft decision circuit 60, where it is determined whether the data is within the range of the upper II set value αCjk and the lower limit set value Cjk retrieved from the set value storage circuit 6s. A correlation-order determination is made as to whether or not the signal is within the 1 range, logic 11. If it is outside the range, logic 0 is output. Such correlation-order determination processing is performed for all specified combinations of window areas.

Next, in the correlation secondary determination circuit 61, the correlation from the one circuit 60 -
The correlation between the next determination result and the set value storage circuit 65 -
By comparing and checking with the next judgment table, a second correlation judgment is performed. If they match, the judgment for the feature item of the group is good.◎ 1113 Window area set as shown in Figure FIG. 116 shows an example of a correlation-order determination table obtained when the number 8 is slightly shifted upward from the position relative to the window area shown in FIG. 20.

In Figure 116, pattern 8 has shifted slightly upward, so
If the number of features extracted by 21 alone is small, one data is a theory I.
Unable to become a II, he remains in the discussion mode,
On the other hand, it will be understood that the sum of the feature amounts of the window areas 21 and 31 finally makes the data ``Logic III''. The same can be said of the window areas 24% and 27. The correlation-order judgment table is also obtained in advance and stored in the setting value storage file 165 by the board 66, which is the same as the case of the first-order judgment table explained earlier. A secondary judgment process is performed using feature data directly measured across the area, and a correlation secondary judgment circuit is performed using correlation calculation data obtained by processing the measurement data. % General Meeting Judgment
1682 outputs a determination result that the product is non-defective.

11711 - 1 -1
1881 shows the details of each operation as a flowchart.
Note that in the eighth section, the note window area is simply referred to as a window.◎ In the explanation of the embodiments described above, the shape of the electrical area is assumed to be rectangular. , circular, annular, etc., in any shape.1 Especially when the receiving pattern is E-shaped rather than a letter, it is important to select a window area with an appropriate shape. It is.

In addition, in the description of the embodiment, the correlation determination process is performed by S
Although it was explained that only the interphase-order judgment results are used during the % correlation secondary judgment, in addition to the correlation-order judgment results, -
It may be done by referring to the next judgment result 01! Although we have explained the stripping of addition as a correlation operation, depending on the pattern conditions, there are cases where it may be advantageous to take 4 [ or a ratio. In some cases, it may be advantageous to correlate different feature items with each other. Operation 71 in such a case: Show the I-chart at @91iH 0 In the above explanation, flush and boundary length were explained as feature items, but the points of intersection, end points, arcuate points, and oblique Is- i!
As mentioned above, this device uses %
The shape of the window area, position, enemy conditions, various threshold values, correlation conditions, various base judgment conditions, etc. can be freely set in the memory fence using the blue electric pattern by keyboard input. Therefore, conventional nositane inspection equipment 1t41 had different hardware and algorithms depending on the target pattern, and it was difficult to mass-produce, but this equipment has a significantly high adaptability to pattern changes that require increased inspection power. It has a configuration that can be mass-produced.Next, the effects of the present invention will be described. It is possible to meet the setting even in areas where no pattern exists, which was previously impossible.
Pattern +α] can also be easily determined to be defective.

■ Judgments in each window area are made using independent upper and lower threshold thresholds for each feature item and each window area, so it is possible to check not only the presence or absence of a pattern within the area, but also changes in the size of the pattern and changes in shadow shape. can be determined quantitatively, and the determination accuracy can be significantly improved even when the pattern is slightly distorted.

Furthermore, even if the density of the power increase pattern differs depending on the region, the test sugar content can be determined according to the pattern by adjusting the threshold value.

■ Window areas can be grouped and each group can be judged independently, so even when there are variations (minor positional deviations or density changes) in good samples, there is no waste spring (mistaking a good product as a defective one). ) will decrease.

■ By using not only measured data to determine whether a pattern is good or bad, but also correlation calculation data between various window areas, patterns can be inspected with high precision and stability. There are cases in which the absolute value of the measured value changes relatively often depending on the location, transportation conditions, etc. In such cases, the method of comparing the measured value using the upper and lower limit values may result in too much waste, or the defect detection rate may be too high. However, 1 Even under such conditions, there are cases where ``the phase debris ratio of the measured values of different window areas is constant'', ``the sum of the measured values of different window areas is constant'', or ``the phase debris ratio of the measured values of different window areas is constant'', Such a condition i
″：：It is possible to conduct high-precision inspections for oxidation.゛
□■ When performing a secondary correlation judgment, we not only focus only on the result of the 1st match-up judgment with that of a non-defective product, but also, as mentioned earlier, perform the primary judgment of a non-defective product under the conditions where the correlation judgment is performed. If the results are also checked in combination, the judgment will be made even more strictly. In particular, even if the target sample is misaligned, the pattern can be inspected with high precision without wasting time.

In the case of 01N, you may be able to find new effects by examining not only the same feature but also the correlation with other feature data (may be the same window area).For example, if dirt If you take the correlation between the direction and the circumference as in an inspection, the difference between good and bad becomes clear.◎ ■ By making the shape of the window area virtually arbitrary, it can be applied to a variety of target patterns, and , since the number of window areas can be reduced, high-speed determination can be achieved.

■ Station as measurement feature i1F! By employing I-box information, changes such as minute defects in patterns can be extracted with high precision.

[Brief explanation of drawings]

1st meter (Xun 1st meter is the principle of conventional pattern inspection equipment)
! 1G2Wti (season is an explanatory diagram showing an example of setting a window area, 1112W mountain) is an explanatory diagram of an example of a pattern, 2nd W
J(C) is an explanation of the pattern seen through the palace area wJ% 3rd m is an explanation showing other setting examples of the 5m1t castle wJ note, 4th
wJ is a block diagram showing an embodiment of the present invention, FIG. 5 is an explanation of the first judgment result block, wJ1, and FIG. 6 is an explanatory diagram showing an example of a correlation-order judgment table.
FIG. 8 is a flowchart showing the details of the operation, and 9th Wi is a flowchart showing another operation example. Code explanation 1.1... Pattern, 2-8 River... Sampling point, 20... Target pattern, 21-
47...M, so...Target pattern%5
1...ITV, 1 melody, 52...Binarization - Prime division 1@s 53e53'...Feature extraction-Route, 54.54' Song...Counting Circuit, 55...
・Memorial memory-path, 56...Song-next judgment circuit, 57...
... Secondary judgment circuit, 58... Song competition judgment necessity circuit,
59...Comparison calculation-path, 6o...Song correlation-order judgment 1gI connection-61...Correlation secondary judgment circuit, 62.
・Comprehensive song judgment-Route 163...Window area generation-Route,
64...Control-road, 6! S...Setting value memory - path, 66...Keyboard agent Patent attorney Naoki Namiki ・Agent Patent attorney Kiyoshi Matsuzaki 22- Figure 1 (lL + tb> Figure 2 ( ^l 1111th ,) Figure IG)

Claims (1)

[Scope of Claims] l) A two-dimensional successive scanning photoelectric conversion device that scans an optical image of a pattern to be inspected and outputs a time-series electrical signal; and (1) divides the electrical signal obtained from the conversion device into pixels. 1) means for setting a window area of *t- within the visual range of the charging conversion device; means for measuring the feature quantities of the pattern to be inspected; and a primary judgment for determining the acceptability of the pattern to be inspected within each captured area based on whether each of the measured feature quantities is within a predetermined threshold range. means and
The window area is grouped into groups, and for each group, a secondary judgment means is configured to judge whether the window area belongs by comparing and comparing it with a preset primary judgment table. 2. In the pattern inspection device according to claim 1, the pattern inspection device is characterized in that the quality of the inspection layer pattern is judged from the secondary judgment result. 1. Means for performing a correlation calculation between the feature quantities measured for each sub-region between a plurality of window regions; 1. Correlation-order determination based on whether or not the correlation calculation result obtained in this manner is within a predetermined threshold range. and means for performing a second correlation judgment by comparing and collating a set of first correlation-order judgment results with a preset correlation-order judgment table; A pattern inspection device characterized in that, when the secondary judgment result is negative, the correlation calculation/judgment means is started and the quality of the pattern to be inspected is judged from the result of the phase-to-phase secondary judgment.