JPH02153480A - Pattern recognizing device - Google Patents

Abstract

PURPOSE:To contrive the improvement of a recognition rate by setting correction data against image pickup data at every split area where a variance is small, and executing the correction of the image pickup data at every split area. CONSTITUTION:Based on correction data determined at every split area of a pattern recognized area 11 of a semiconductor wafer 10, image pickup data from its split area is corrected. As for an illuminance, the variance is small in the split area, therefore, an appropriate correction is executed against data in the split area. Accordingly, even if there is the variance of the illuminance partially at the time of viewing the whole recognition area, the correction corresponding to its split area is executed at every split area. In such a manner, the recognition rate is improved.

Description

[Detailed description of the invention] [Industrial application field]

The present invention relates to a pattern 32m device.

[Conventional technology]

For example, in a semiconductor manufacturing process, a pattern attached to a predetermined part of a workpiece, such as a semiconductor wafer, is imaged, the output is pattern recognized, and the type of workpiece can be confirmed based on the recognition information. In some cases, alignment with the processing device is performed. The pattern recognition device used in this case includes a light source that irradiates a part of the object to be recognized, such as a pattern engraved with a product number, etc., engraved by a laser or the like on a semiconductor wafer, and an imaging device that takes an image of the irradiated pattern engraved part. and a recognition means for pattern recognition of the imaging output. In the recognition means, the imaging data is, for example, binarized, and this binarized pattern information is compared with previously stored pattern information, and a match between the two patterns is taken as a recognition result of the imaging pattern. In this case, the threshold value for binarizing the imaged data has conventionally been set to a predetermined value for the entire pattern recognition area.

[Problems to be Solved by the Invention] By the way, when measuring light is irradiated from a light source to a pattern recognition area of a recognition object, it is almost impossible to uniformly irradiate the entire pattern recognition area. Therefore, the imaging data includes offset values that vary depending on the light irradiation position. If the threshold value for binarization of imaged data is constant throughout the pattern recognition area, the optimal binarization threshold value setting will be difficult because the offset value will vary depending on the light irradiation position. This has the disadvantage that the recognition rate decreases as a result. Moreover, when a diffusely reflective thin film, such as an Aβ film, is formed on the object to be recognized, such as a semiconductor wafer, the shading between the substrate and the pattern is different, making it difficult to select the binarization threshold value. This becomes even more difficult, which also causes a decrease in the recognition rate. This phenomenon is
Even with the naked eye, there is such a lack of contrast that it is difficult to read unless the angle of the illumination light is set in a specific narrow area. Furthermore, even if the binarization threshold value can be determined appropriately, the objects to be recognized may be of different types, such as a semiconductor wafer before an Afl film is formed and a semiconductor tool after an Afl film is formed. , the illuminance conditions within the recognition area of the object to be recognized change, resulting in a decrease in the recognition rate. In view of the above points, the present invention aims to provide a pattern recognition device that can improve the recognition rate even with information without contrast.

[Means for Solving the Problem] In the invention described in claim (1), a light source that illuminates an object to be recognized, an imaging device that images the illuminated object to be recognized, and imaging data from this imaging device are provided. In a pattern recognition device equipped with a recognition unit that performs pattern recognition based on Correction data is set in advance, the imaging data is corrected based on the set correction data, and the corrected data is supplied to the recognition section. In the invention described in claim (2), an illuminance variable means for varying the illuminance on the object to be recognized is provided, and each time the object to be recognized is changed, the maximum value of the illuminance at which the pattern of the object to be recognized can be recognized is changed. The illuminance is determined from the minimum value and the minimum value. In the invention described in claim (3), an illuminance variable means for varying the illuminance on the object to be recognized is provided, and each recognition result when the same object to be recognized is recognized by changing the illuminance is overlapped. At the same time, a recognition result for the object to be recognized is obtained. [Operation] In the invention of claim (1), based on the correction data determined for each divided area of the pattern recognition area, the imaged data from the divided area is corrected. Since the illuminance has small variations within the divided areas, appropriate correction is made to the data within the divided areas. Therefore, even if there is variation in illuminance in parts of the entire recognition area, the recognition rate is improved because correction is made for each divided area in accordance with that divided area. In the invention of claim (2), even if the object to be recognized changes and the illuminance conditions change, the illuminance will be adjusted according to each object to be recognized, so an improvement in the recognition rate can be expected. In the invention of claim (3), since the recognition results are superimposed, even if there is a pattern that cannot be recognized at a certain illuminance, if it is recognized at another illuminance, it becomes possible to recognize all the patterns as a whole. be.

【Example】

Hereinafter, one embodiment of the present invention will be described with reference to the drawings. In FIG. 1, reference numeral 10 denotes a semiconductor wafer as an example of an object to be recognized, and in a region 11 near the orientation flat of this semiconductor wafer 10, as shown in FIG. 2, a character pattern is engraved with a laser beam. It is formed. Reference numeral 12 denotes a light source from which a region 11 of the semiconductor wafer 10 is irradiated with measurement light consisting of parallel light or weakly focused light. 13 is an imaging device, for example a COD camera. In this case, the CCD camera 13 is arranged in the direction of the reflected optical axis, which is slightly shifted from the optical axis direction a of the specularly reflected light of the measurement light irradiated from the light source 12 onto the region 11 of the semiconductor wafer 10. With this arrangement, the object to be recognized is the semiconductor wafer 1.
0) Images with good contrast can be obtained even when there is a lot of diffuse reflection on the surface (Special IF related to the applicant's application)
J No. 63-79062 and patent application (2) dated July 19, 1988). The imaging output signal from the CCD camera 13 is supplied to the A/D converter 15 via the amplifier 14, and is exchanged into digital data of, for example, about 8 bits. This digital data is supplied to filtering circuit 16. This filtering circuit 16 is supplied with filter coefficients from a filter coefficient table 17, and blurs,
Filtering processing such as shrinkage and noise removal is performed. In this case, the filter coefficient table 17 has written coefficients corresponding to the coordinates of each imaging position when the area 11 on the surface of the semiconductor wafer 10 to be pattern recognized is considered as a two-dimensional plane. The output of the filtering circuit 16 is supplied to a binarization circuit 18. This binarization circuit 18 includes a threshold table 1
A binarization threshold value is supplied from 9. In the case of this example, the threshold value table 19 includes the area 11 to be pattern recognized divided into a plurality of areas as shown in FIG. The threshold value Tij(i,
j=1.2,3°4) is written in advance. For example, when region 11 is divided into 16, 1 from T11 to T44
Sixteen appropriate threshold values are written, each corresponding to a 6-minute vJ region. The same threshold table 19 is used for semiconductor wafers of the same type, and when the type of semiconductor wafer as the object to be recognized is changed, the threshold value table 19 is used for the recognition area 11 of the new semiconductor wafer. The threshold value is rewritten according to each divided area. Of course, if the lot of semiconductor wafers of the same type changes and a change is desired, of course,
In this case, in order to prevent the threshold value as a correction value from suddenly changing greatly at the boundary of each divided area, the threshold value near the boundary of each divided area should be changed to the value of the adjacent divided area. Smoothing is performed so that the area gradually approaches the threshold value. This processing can be performed using a digital filter. Note that instead of changing the binarization threshold value,
An offset value corresponding to the brightness (illuminance) of each divided area may be added to the data before being converted into values. This offset value can be included in the filter coefficient . Therefore, when adding the offset value, the filter coefficient in the filter coefficient table 17 may include an offset value corresponding to each lq area. In this case as well, the so-called smoothing process of the correction values at the boundaries between the divided areas is performed in the same manner as described above. Data from the binarization circuit 18 is supplied to a character extraction circuit 20. This character cutting circuit 20 sequentially obtains pattern data for each unit to be pattern recognized, which is supplied to a pattern matching circuit 21 and compared with character pattern data stored in a pattern memory 22, and the comparison is performed. A recognition output corresponding to the result is obtained from the recognition output circuit 23 and output to the output terminal 24. As described above, in the present invention, the pattern recognition area is divided into a plurality of areas, and the imaged data is corrected using the binarization threshold value set for each divided area and the offset value for the imaged data. When looking at the entire pattern recognition area, even if there is variation in brightness in parts, the variation in brightness is reduced within the divided areas, so it is possible to improve the recognition rate. In this example, the amount of light from the light source 12 can be controlled so that the illuminance (integrated value over the entire area 11) in the pattern recognition area 11 of the semiconductor wafer 10 is appropriate. This amount of light is generally set to the same value if the semiconductor wafers to be recognized are of the same type, and is changed to an appropriate value each time the type of semiconductor wafer is changed. For this reason, in this example, the light source 12 is a variable light source and is provided with a light amount control means 25, so that when the type of semiconductor wafer that is the object to be recognized is changed,
Scanning is performed in which the amount of light is changed at a constant step width, and an appropriate value is set based on the result. Including this light amount control means 25, in this example, A
Data processing after /D conversion can be configured by a microcomputer. The light amount control process will be explained below with reference to the flowchart shown in FIG. In this example, scanning is started from a predetermined minimum light amount of 5 Hin, and the predetermined light amount changing step 5
Change the light intensity sequentially with Sca, and reach the specified maximum light intensity of 5Ha.
End scanning with x. First, the current light intensity [■] of the light source 12 is set to the minimum light intensity of 5 Hin for the scanning width (step 101), the pattern in the area 11 of the semiconductor wafer 10 is read using this light intensity (step 102), and the reading state is determined. is performed (step 103), this step 103
If it is determined that all the patterns can be read, the process proceeds to step 104, where it is determined whether this is the first time that all the patterns can be read. If it is the first time, the process proceeds to step 105, where the minimum light amount LIHin that allows all patterns to be read and the maximum light amount LIHax that allows all patterns to be read are set to the current light amount [■]. If it is not the first time, the process proceeds to step 106, and the maximum light amount LIHax for reading all patterns is set to the current light amount at that time. As a result of the determination in step 103, if the entire pattern could not be read but the character pattern could be extracted, the process proceeds to step 107, where it is determined whether or not this is the first time that only that extraction became possible. be done. If it is the first time, proceed to step 108 and select the minimum amount of light that can be extracted [2
Let the old n and the maximum light amount L2Hax that can be cut out be the current light amount LT at that time. If it is not the first time, the process proceeds to step 109, and the maximum light amount L2Max that can be extracted is set to the current light amount at that time. After steps 105, 106, 108 or 109, the process proceeds to step 111. If it is determined in step 103 that neither reading nor cropping is possible at all, the process proceeds to step 110, where it is determined whether there is an effective light amount [1] for all patterns or the light amount [2] that can be extracted. If it is determined that neither is valid, the process advances to step 111. In step 111, the current light amount LT is increased by the amount of light amount change step 5Sca. Next, the process proceeds to step 112, where it is determined whether the increased current light iLT is larger than the maximum light amount 5Hax of the scanning width. The process returns to step 102, and the above operations are repeated. When the current light intensity [■ becomes larger than the maximum light intensity 5Hax, scanning stops and step 1
Proceed to step 13. When scanning stops due to the above operations,
Within the scanning width, the minimum light amount that allows all patterns to be read [1 old n, the maximum light amount that allows all patterns to be read LIHax
, the minimum amount of light that can be extracted L2Hin, and the maximum amount of light that can be extracted L2Hax are determined. Further, when it is determined in step 110 that reading and cropping are not possible but there is a valid value, either the total pattern readable light amount [1 or the cropping possible light amount E2], the process proceeds to step 113. In this step 113, the total pattern readable light amount [
1 is valid or not. If valid, proceed to step 114 and calculate (LIHax + LIHin)'-2=cpi
Alternatively, the light amount Lpl is determined as m+n (m, n are predetermined internal division ratios), the current light JiLT is set to this light amount Lpl, and the program ends. When it is determined in step 113 that the total pattern readable light amount L1 is not valid, the process proceeds to step 115, where it is determined whether the cutout possible light amount [2 is valid. Then, if it is determined to be valid, step 1
Proceeding to step 16, the light amount 1112 (L2Hax +L2Hin)-2=Lp2 or m+n (m, n are predetermined internal division ratios) is obtained, the current light amount [■ is set as this light amount Lp2, and the program ends. . If it is determined in step 115 that the extractable light amount [2 is not valid, it is determined that an error has occurred (step 117). As described above, the light intensity [■] of the light source 12 is determined to be an appropriate light intensity for each type of semiconductor wafer. In this way, the light intensity of the light source 12 is adjusted appropriately, the pattern recognition area is divided into a plurality of areas, and each divided area is separated. In consideration of the case where a part of the pattern in the pattern recognition area cannot be read even if the i-image data is corrected, the following is further performed in this example. In other words, the light intensity of the light source 12 is changed at a predetermined light intensity change pitch of 5 Pit within a predetermined scan width ±Sld centered on the light ff1LT set as described above, and the recognition results for each light intensity are superimposed. To enable reading of all patterns in the pattern recognition area. The automatic reading algorithm will be explained below with reference to the flowchart of FIG. In the explanation of FIG. 4, lO is the initial light intensity of the light source 12, ρ
is the current light intensity, I Hin=j! 0-514id, j
! Ha × = RO + 5 Wid, 6, t, s ig
h is the sign bit, and when it is +1, the current light amount β
When this is -1, the current light amount l is scanned in a direction smaller than the initial light amount 10. Further, Hlok is an identification bit indicating whether or not scanning in a large direction is possible, and when Hiok=1, it means that scanning is possible. LOOk is an identification bit indicating whether scanning in the small direction is possible or not, and when LOok=1, it means that scanning is possible. First, in step 201, initialization is performed. At this time, for example, the sign bit is high, the identification bit is old Ok, and [00
Then, in step 202, the state of the sign bit "sigh" is determined. If sigh=1, the process proceeds to steps 301 to 309 for scanning in the large direction, and if sigh=o, the process proceeds to steps 401 to 408 for scanning in the small direction. In scanning in a large direction, first, in step 301, the number of times n of light intensity changes is increased by one. Next, step 302
It is determined whether the identification bit tllok is 1 or not. If the identification bit old Ok is not 1, the process proceeds to step 311, where it is determined whether the identification bit [00k is 1 or not, and the 1
If so, proceed to step 402 of scanning in the small direction,
If the identification bit [00k is not 1, it is considered an error and the process ends. If it is determined in step 302 that the identification bit old Ok is 1, the process proceeds to step 303, where the light amount l is the initial light amount l.
It is increased by 5Pitxn from O. Next, in step 304, the light amount l is the maximum light amount j!
It is determined whether Hax has been exceeded. If it exceeds, the process proceeds to step 310, where the identification bit is set to old gh.0, and then the process proceeds to step 311, where scanning in the major direction is stopped. If not, proceed to step 305,
A character string to be pattern recognized is read. Then, the process proceeds to step 306, where it is determined whether or not the character can be extracted. If it is possible, the process proceeds to step 307, where it is superimposed on the previously read character string, and in step 308, whether the character string is completed or not. In other words, it is determined whether all characters have been recognized. If the character string is not completed, the sign bit is set to -1 in step 309, and then the process returns to step 202. If it is determined in step 308 that the character string is complete, the program ends. Next, in the scan in the small direction, first, step 4
01, it is determined whether the identification bit 100k is 1 or not. If the identification bit Look is not 1, step 410
It is determined whether the identification bit old OK is 1 or not.
If it is 1, the process proceeds to step 302 for scanning in the larger direction, and if the identification bit old Ok is not 1, it is determined that there is an error and the process ends. If it is determined in step 401 that the identification bit [00k is 1, the process proceeds to step 402, where the light amount l is decreased from the initial light amount 10 by 5Pitxn. Next, in step 403, the light il is the minimum light amount j!
It is determined whether it is equal to or greater than Min. Minimum light amount j!
If it is less than Min, the process proceeds to step 409, where the identification bit LOok is set to O, and then the process proceeds to step 410, where scanning in the small direction is stopped. Fi low light amount j! Hen
If this is the case, the process proceeds to step 404, where the character string to be pattern recognized is read. And step 40
In step 5, it is determined whether or not it is possible to cut out the character, and if so, the process advances to step 406, where it is superimposed on the previously read character string, and in step 407, whether the character string is completed or not. , that is, it is determined whether all are recognized or not,
If the character string is not completed, the sign bit is set to 1 in step 408, and then the process returns to step 202. If it is determined in step 407 that the character string is complete, the program ends. In the above example, the amount of light from the light source 12 is changed as a means of changing the illuminance of the recognition area of the object to be recognized, but even if the distance between the light source 12 and the object to be recognized is changed, the illuminance will not change. changes can be made. Further, the above example is a case of recognizing a character pattern engraved on a semiconductor wafer, but the present invention is applicable to the case where the object to be recognized is irradiated with light and imaged. It can be of any kind. For example, FA parts,
It goes without saying that this can be applied to tools, and even to characters written on paper. Further, the present invention is applicable not only to character patterns but also to graphic pattern and other various pattern recognition. [Effects of the Invention] Even if the variation in illuminance is large in the entire pattern recognition area of the object to be recognized, the variation becomes small within each divided area into which this recognition area is divided. According to this invention, the correction data for the imaging data is set for each divided region with small variations, and the imaging data is corrected for each divided region, so that the data within the divided region is corrected appropriately. Corrections are made. Therefore, when looking at the entire recognition area, even if there are variations in parts,
Since correction is performed for each divided area according to the divided area, the recognition rate is improved. Moreover, since the illuminance of the object to be recognized can be changed to an appropriate value each time the object to be recognized changes, this also contributes to improving the recognition rate. Furthermore, in this invention, it is possible to change the illuminance and superimpose the recognition results at each illuminance to obtain the overall recognition result. If it can be recognized, all patterns can be recognized as a whole.

[Brief explanation of the drawing]

FIG. 1 is a block diagram of an embodiment of the present invention, FIG. 2 is a diagram showing an example of an object to be recognized, and FIGS. 3 and 4 are flowcharts for explaining the invention in detail. be. Semiconductor wafer as an example of object to be recognized Light source imaging device A/D converter Binarization circuit Threshold table Light amount variable means

Claims (3)

[Claims] (1) A pattern recognition device that includes a light source that illuminates an object to be recognized, an imaging device that images the illuminated object to be recognized, and a recognition section that performs pattern recognition based on imaging data from the imaging device. The pattern recognition area of the recognition object is divided into a plurality of divided areas, and correction data corresponding to the divided area is set in advance for the imaging data from each divided area, and based on the set correction data. A pattern recognition device in which the imaging data is corrected and the corrected data is supplied to the recognition unit. (2) A pattern recognition device that includes a light source that illuminates the object to be recognized, an imaging device that images the illuminated object to be recognized, and a recognition section that performs pattern recognition based on the imaging data from the imaging device. An illuminance variable means for varying the illuminance on the recognition object is provided, and each time the recognition object is changed, the illuminance is determined from the maximum value and the minimum value of the illuminance at which the pattern of the recognition object can be recognized. pattern recognition device. (3) A pattern recognition device comprising a light source that illuminates the object to be recognized, an imaging device that images the illuminated object to be recognized, and a recognition unit that performs pattern recognition based on the imaging data from the imaging device. An illuminance variable means for varying the illuminance on the recognition object is provided, and the recognition results for the same recognition object obtained by changing the illuminance are superimposed to obtain the recognition result for the recognition object. A pattern recognition device.