CN117974601A - Method and system for detecting surface defects of silicon wafer based on template matching - Google Patents
Method and system for detecting surface defects of silicon wafer based on template matching Download PDFInfo
- Publication number
- CN117974601A CN117974601A CN202410139542.0A CN202410139542A CN117974601A CN 117974601 A CN117974601 A CN 117974601A CN 202410139542 A CN202410139542 A CN 202410139542A CN 117974601 A CN117974601 A CN 117974601A
- Authority
- CN
- China
- Prior art keywords
- silicon wafer
- image
- template
- matching
- difference model
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- XUIMIQQOPSSXEZ-UHFFFAOYSA-N Silicon Chemical compound [Si] XUIMIQQOPSSXEZ-UHFFFAOYSA-N 0.000 title claims abstract description 160
- 229910052710 silicon Inorganic materials 0.000 title claims abstract description 160
- 239000010703 silicon Substances 0.000 title claims abstract description 160
- 230000007547 defect Effects 0.000 title claims abstract description 73
- 238000000034 method Methods 0.000 title claims abstract description 50
- 238000012549 training Methods 0.000 claims abstract description 52
- 238000007689 inspection Methods 0.000 claims abstract description 45
- 230000011218 segmentation Effects 0.000 claims abstract description 33
- 238000001514 detection method Methods 0.000 claims abstract description 18
- 238000001914 filtration Methods 0.000 claims description 16
- 238000013519 translation Methods 0.000 claims description 14
- 238000011176 pooling Methods 0.000 claims description 7
- 230000008859 change Effects 0.000 claims description 4
- 230000006870 function Effects 0.000 claims description 4
- 239000011159 matrix material Substances 0.000 claims description 4
- 230000007246 mechanism Effects 0.000 claims description 4
- PXFBZOLANLWPMH-UHFFFAOYSA-N 16-Epiaffinine Natural products C1C(C2=CC=CC=C2N2)=C2C(=O)CC2C(=CC)CN(C)C1C2CO PXFBZOLANLWPMH-UHFFFAOYSA-N 0.000 claims description 3
- 230000003213 activating effect Effects 0.000 claims description 3
- 230000004913 activation Effects 0.000 claims description 3
- 238000012937 correction Methods 0.000 claims description 3
- 238000003708 edge detection Methods 0.000 claims description 3
- 238000005070 sampling Methods 0.000 claims description 3
- 230000001502 supplementing effect Effects 0.000 claims description 3
- 230000009466 transformation Effects 0.000 claims description 3
- 235000012431 wafers Nutrition 0.000 abstract description 111
- 230000008569 process Effects 0.000 abstract description 7
- 238000004519 manufacturing process Methods 0.000 description 17
- 238000003860 storage Methods 0.000 description 9
- 230000002950 deficient Effects 0.000 description 4
- 238000012216 screening Methods 0.000 description 3
- 238000010586 diagram Methods 0.000 description 2
- 238000009826 distribution Methods 0.000 description 2
- 238000005286 illumination Methods 0.000 description 2
- 230000003287 optical effect Effects 0.000 description 2
- 238000012545 processing Methods 0.000 description 2
- 230000004075 alteration Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000005540 biological transmission Effects 0.000 description 1
- 238000005336 cracking Methods 0.000 description 1
- 238000003745 diagnosis Methods 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 238000011156 evaluation Methods 0.000 description 1
- 238000009499 grossing Methods 0.000 description 1
- 239000000463 material Substances 0.000 description 1
- 238000012544 monitoring process Methods 0.000 description 1
- 238000005457 optimization Methods 0.000 description 1
- 238000003909 pattern recognition Methods 0.000 description 1
- 239000002994 raw material Substances 0.000 description 1
- 230000004044 response Effects 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 230000003313 weakening effect Effects 0.000 description 1
Landscapes
- Testing Or Measuring Of Semiconductors Or The Like (AREA)
- Investigating Materials By The Use Of Optical Means Adapted For Particular Applications (AREA)
Abstract
The invention discloses a method and a system for detecting surface defects of a silicon wafer based on template matching, comprising the steps of collecting images of good silicon wafers, and carrying out automatic global threshold segmentation and rectangular segmentation to obtain a region of interest; pyramid operation is carried out on the filtered image, a corresponding shape outline is created, and a difference model is built; training the difference model by using a training image of a standard silicon wafer, and registering an inspection template in the difference model; and acquiring a silicon wafer image to be detected, extracting an interested region, guiding the interested region of the silicon wafer image to be detected into a difference model, matching by using an inspection template, and extracting a matching difference according to a matching result. The invention utilizes the image pyramid enhancement template matching process, realizes more robust and accurate matching through multi-scale matching, and provides accurate positioning of defects, thereby improving the reliability of feature detection or defect identification.
Description
Technical Field
The invention relates to the technical field of solar photovoltaic cell manufacturing, in particular to a method and a system for detecting surface defects of a silicon wafer based on template matching.
Background
Silicon wafer production is an important link in the production of solar photovoltaic cells. During the production process of the silicon wafer, due to the influences of factors such as manufacturing technology, materials, production equipment and the like, the defects such as chromatic aberration, hidden cracking, edge breakage, unfilled corner, tiny notch and the like often exist. Such defects can lead to product quality of subsequent solar photovoltaic cells not reaching standards. In order to ensure the yield of silicon wafer production, the defects are required to be detected and screened out.
The traditional detection method adopts a human eye mode for detection, is low in accuracy and efficiency, and has the conditions of false detection and missing detection. And the silicon chip is easy to pollute by manual operation. Therefore, how to timely and accurately screen and detect defects such as edge breakage and unfilled corner of the silicon wafer in a template matching mode, help to improve the silicon wafer production flow, and improve the product quality and the production efficiency is one of the problems to be solved urgently.
Disclosure of Invention
In order to solve the technical problems, the invention provides a complex equipment fault diagnosis method and system based on multidimensional features, which are used for solving the problem of low efficiency of the existing artificial silicon wafer defect detection mode.
The first aspect of the invention provides a method for detecting surface defects of a silicon wafer based on template matching, which comprises the following steps:
acquiring good silicon wafer images, performing automatic global threshold segmentation and rectangular segmentation on the good silicon wafer images, removing silicon wafer background, and acquiring an interested region;
Performing mean filtering on the region of interest, performing pyramid operation on the filtered image, creating a corresponding shape outline, and constructing a difference model;
Training the difference model by using a training image of a standard silicon wafer, registering the inspection templates in the difference model, and optimizing the corresponding inspection templates;
the method comprises the steps of obtaining a silicon wafer image to be detected, extracting an interested region, guiding the interested region of the silicon wafer image to be detected into a difference model, utilizing an inspection template to match, extracting matching differences according to a matching result, and visually displaying the difference part.
In this scheme, carry out automatic global threshold segmentation and rectangle segmentation to the good silicon chip image, specifically do:
Acquiring good silicon wafer images, dividing the good silicon wafer images into a background area and a foreground area according to pixel intensity values, and automatically determining an optimal threshold value by adopting an OSTU algorithm;
carrying out automatic global threshold segmentation on the good silicon wafer image through the optimal threshold value, and acquiring a placement area of the good silicon wafer in a foreground area;
And obtaining edge characteristics in the placement area of the good silicon wafer by using a Canny edge detection algorithm, performing rectangular segmentation based on the edge characteristics, and further segmenting the background to obtain an interested area.
In the scheme, mean value filtering is carried out on the region of interest, pyramid operation is carried out on the filtered image, a corresponding shape outline is created, and a difference model is built, specifically:
The method comprises the steps of carrying out mean filtering on the region of interest, reducing high-frequency noise through the change of smooth pixel intensity, extracting original multi-scale features of the region of interest by utilizing an image pyramid, and embedding a channel attention mechanism to filter the original multi-scale features to obtain filtered multi-scale features;
Performing convolution operation on the multi-scale features to obtain salient features, obtaining salient objects with different sizes and shapes, and splicing the salient features with the same resolution in the original multi-scale features after feature up-sampling;
Fully fusing the spliced features by utilizing convolution operation, supplementing the space detail features of the salient objects, and creating the shape outline of the silicon wafer in the region of interest;
And constructing the difference model by taking the obtained shape outline as a template image of the difference model and taking an original image of a standard silicon wafer as input.
In this scheme, training the difference model by using a training image of a standard silicon wafer, registering an inspection template in the difference model, and optimizing the corresponding inspection template, specifically:
obtaining an original image of a standard silicon wafer as a training image of a difference model, carrying out local division on the training image and a template image of an inspection template in the difference model, carrying out local feature coding in a local area, and setting two rotation and translation learning training paths in the difference model;
coding single pixel points in a local area by utilizing a multi-layer sensing module to obtain local single point characteristics, carrying out average pooling on the local single point characteristics to obtain local integral characteristics of the local area, and splicing the local single point characteristics and the local integral characteristics in characteristic channel dimensions;
activating the spliced features by using an activation function, performing batch standardization and maximum pooling operation, outputting final local features, and acquiring rotation and translation information of a training image and a template image according to the final local features in two learning training paths;
Obtaining attention weights according to the similarity between the training image and the local area of the template image in the rotation and translation information, weighting the local area of the training image by using the attention weights, and adding the weighted local area of the training image to the template image through residual errors;
And acquiring optimized template images in the rotation and translation two learning training paths, and optimizing an inspection template in the difference model.
In the scheme, an optimized inspection template in a difference model is obtained, gray value parameters of the difference model are determined, and a maximum gray threshold value and a minimum gray threshold value are calculated according to the gray value parameters;
And obtaining a maximum gray value image and a minimum gray value image corresponding to the silicon chip according to the maximum gray threshold value and the minimum gray threshold value, and storing the maximum gray value image and the minimum gray value image into the difference model.
In the scheme, a silicon wafer image to be detected is obtained to extract an interested region, the interested region of the silicon wafer image to be detected is led into a difference model, and an inspection template is utilized for matching, specifically:
acquiring a silicon wafer image to be detected by using a silicon wafer image acquisition device, and performing automatic global threshold segmentation and rectangular segmentation on the silicon wafer image to be detected to acquire an interested region;
Introducing a region of interest into a difference model to extract edge characteristics of a silicon wafer, calculating similarity between an image to be detected and a detection template according to normalized product correlation by utilizing the edge characteristics, performing hierarchical search and matching in an image pyramid according to the similarity, and performing matching from the pyramid tip;
searching surrounding local areas in the next layer of the image pyramid according to the matching position of the pyramid tip, continuing to match, and acquiring areas with similarity meeting preset standards from the corresponding search image of the silicon wafer image to be detected, so as to finish matching and positioning of the silicon wafer image to be detected;
and carrying out affine transformation by using the homogeneous rotation matrix after matching and positioning, and carrying out position correction.
In the scheme, the matching difference is extracted according to the matching result, and the difference part is visually displayed, specifically:
obtaining an average gray value of a region of interest in a silicon wafer image to be detected, and scaling the average gray value of the region of interest according to an inspection template of a difference model;
And comparing the scaled average gray value with the maximum gray value image and the minimum gray value image in the difference model, outputting a silicon wafer defect area according to a comparison result, and visually marking the exact position of the defect area.
The second aspect of the invention also provides a system for detecting the surface defects of the silicon wafer based on template matching, which comprises: the device comprises a memory and a processor, wherein the memory comprises a silicon wafer surface defect detection method program based on template matching, and the silicon wafer surface defect detection method program based on the template matching realizes the following steps when being executed by the processor:
acquiring good silicon wafer images, performing automatic global threshold segmentation and rectangular segmentation on the good silicon wafer images, removing silicon wafer background, and acquiring an interested region;
Performing mean filtering on the region of interest, performing pyramid operation on the filtered image, creating a corresponding shape outline, and constructing a difference model;
Training the difference model by using a training image of a standard silicon wafer, registering the inspection templates in the difference model, and optimizing the corresponding inspection templates;
the method comprises the steps of obtaining a silicon wafer image to be detected, extracting an interested region, guiding the interested region of the silicon wafer image to be detected into a difference model, utilizing an inspection template to match, extracting matching differences according to a matching result, and visually displaying the difference part.
The invention discloses a method and a system for detecting surface defects of a silicon wafer based on template matching, comprising the steps of collecting images of good silicon wafers, and carrying out automatic global threshold segmentation and rectangular segmentation to obtain a region of interest; pyramid operation is carried out on the filtered image, a corresponding shape outline is created, and a difference model is built; training the difference model by using a training image of a standard silicon wafer, and registering an inspection template in the difference model; and acquiring a silicon wafer image to be detected, extracting an interested region, guiding the interested region of the silicon wafer image to be detected into a difference model, matching by using an inspection template, and extracting a matching difference according to a matching result. The invention utilizes the image pyramid enhancement template matching process, realizes more robust and accurate matching through multi-scale matching, and provides accurate positioning of defects, thereby improving the reliability of feature detection or defect identification.
Drawings
FIG. 1 shows a flow chart of a method for detecting surface defects of a silicon wafer based on template matching;
FIG. 2 is a flow chart of a method for acquiring the shape profile of a silicon wafer image according to the present invention;
FIG. 3 is a flow chart of a method of inspecting templates in an optimized disparity model of the present invention;
FIG. 4 shows a block diagram of a template matching based silicon wafer surface defect detection system of the present invention.
Detailed Description
In order that the above-recited objects, features and advantages of the present application will be more clearly understood, a more particular description of the application will be rendered by reference to the appended drawings and appended detailed description. It should be noted that, without conflict, the embodiments of the present application and features in the embodiments may be combined with each other.
In the following description, numerous specific details are set forth in order to provide a thorough understanding of the present invention, but the present invention may be practiced in other ways than those described herein, and therefore the scope of the present invention is not limited to the specific embodiments disclosed below.
FIG. 1 shows a flow chart of the method for detecting the surface defects of the silicon wafer based on template matching.
As shown in fig. 1, the first aspect of the present invention provides a method for detecting a surface defect of a silicon wafer based on template matching, which includes:
S102, acquiring good silicon wafer images, performing automatic global threshold segmentation and rectangular segmentation on the good silicon wafer images, removing silicon wafer background, and acquiring an interested region;
S104, carrying out mean value filtering on the region of interest, carrying out pyramid operation on the filtered image, creating a corresponding shape outline, and constructing a difference model;
s106, training the difference model by using a training image of a standard silicon wafer, registering an inspection template in the difference model, and optimizing the corresponding inspection template;
S108, acquiring a silicon wafer image to be detected, extracting an interested region, guiding the interested region of the silicon wafer image to be detected into a difference model, utilizing an inspection template to match, extracting a matching difference according to a matching result, and visually displaying the difference part.
The silicon wafer image acquisition device mainly comprises a large-target-surface high-resolution area array industrial camera, a high-uniformity CCD square backlight source and a silicon wafer transmission belt. Because the silicon chip generates gray level difference with surrounding background under the background of the backlight source, the gray level difference can be utilized to carry out background filtering, firstly, the imported good product image is subjected to automatic global threshold segmentation, firstly, the larger part of the gray level value and other parts with larger difference is extracted, then rectangular segmentation is carried out, the corresponding silicon chip image is segmented from the background containing the belt, only the image part is required to be processed, the good product silicon chip image is divided into a background area and a foreground area according to the pixel intensity value, and the optimal threshold value is automatically determined by adopting an OSTU algorithm; the OSTU algorithm provides an automatic method to determine the optimal threshold without human intervention, which is critical to the silicon wafer inspection process. And the OSTU method can also reduce the condition of continuously adjusting the threshold value according to the change of illumination or other factors, and meet the adaptability requirement of the silicon wafer detection process to different production conditions.
In the implementation process of the OSTU algorithm, assuming that a threshold TH exists to divide all pixels of an image into two classes C1 (< TH) and C2 (> TH), the average value of each of the two classes of pixels is m 1、m2, and the global average value of the image is m G. Meanwhile, the probability that the pixels are divided into C1 and C2 types is p 1、p2 respectively; wherein p 1*m1+p2*m2＝mG,p1+p2 = 1; according to the concept of variance, the expression of the inter-class variance sigma 2 is sigma 2＝p1(m1-mG)2+p2(m2-mG)2, p 1*m1+p2*m2＝mG is substituted into the expression of the inter-class variance, sigma 2＝p1p2(m1-m2)2 can be obtained, the gray level k which maximizes the above expression is obtained to be an OSTU threshold, the optimal threshold is used for carrying out automatic global threshold segmentation on the good silicon wafer image, and the placement area of the good silicon wafer is obtained in the foreground area;
And (3) acquiring edge characteristics in the placement area of the good silicon wafer by using a Canny edge detection algorithm, and calculating specific shape attributes to filter or select the shape once the contour is detected. And (3) generally calculating the attributes such as area, perimeter, mass center, bounding box, length-width ratio, roundness and the like, dividing the rectangle based on the edge characteristics, screening the conditions corresponding to the rectangle, returning a group of new regions with specified conditions according to the selected shape parameters, and obtaining the region of interest.
FIG. 2 shows a flow chart of a method of the present invention for obtaining the shape profile of a silicon wafer image.
According to the embodiment of the invention, the mean value filtering is carried out on the region of interest, pyramid operation is carried out on the filtered image, a corresponding shape outline is created, and a difference model is built, specifically:
S202, carrying out mean value filtering on the region of interest, reducing high-frequency noise through the change of smooth pixel intensity, extracting original multi-scale features of the region of interest by utilizing an image pyramid, and embedding a channel attention mechanism to filter the original multi-scale features to obtain filtered multi-scale features;
S204, performing convolution operation on the multi-scale features to obtain significant features, obtaining significant objects with different sizes and shapes, and splicing the significant features with the same resolution in the original multi-scale features after feature up-sampling;
S206, fully fusing the spliced features by utilizing convolution operation, supplementing the space detail features of the salient objects, and creating a shape contour of the silicon wafer in the region of interest;
s208, taking the obtained shape outline as a template image of the difference model, and constructing the difference model by matching with an original image of a standard silicon wafer as input.
It should be noted that the average filter is generally called a box filter or an average filter, and replaces each pixel in the image with an average value of its neighboring pixels. In silicon wafer inspection, there may be inherent noise or defects in the image acquired by the wafer during inspection. The mean filtering helps to reduce high frequency noise by smoothing variations in pixel intensity. This is particularly effective for improving the visibility of the pattern or feature of interest, enabling the defective portion to be made more distinct from the surrounding in the present invention. The wafer may have slight surface irregularities or variations, and the mean filtering helps to eliminate such variations, enabling detection to be focused on relevant features and defects.
The image pyramid may preferably be a gaussian pyramid or a laplacian pyramid, an image pyramid being a multi-scale representation of an image, wherein each level in the pyramid corresponds to a different scale of the original image. Which provides a scale-space representation of the original image capturing information at multiple scales. The size and scale of features or defects of a silicon wafer may vary, and the pyramid allows for the detection of features of different resolutions, and the image pyramid helps locate features and defects by providing a multi-resolution view of the image. This is particularly useful when inspecting silicon wafers having features of different sizes and helps identify the exact location of the defect. Where template matching or pattern recognition is involved, the image pyramid may enhance the matching process. The multi-scale matching can realize more robust and accurate matching, thereby improving the reliability of feature detection or defect identification.
Embedding a channel attention mechanism in an image pyramid, pooling the features in each channel, acquiring the correlation between the channels by using a full-connection layer, generating attention weights according to the correlation, filtering the features of different channels, regarding the features with higher response degree in the channels as important features, distributing more weights for the important features, and weakening the background information in the silicon wafer image. And carrying out deep convolution on the multi-scale features to read the obvious features corresponding to the multi-scale features, wherein the obvious features often lack space structure features, so that the extracted obvious features are supplemented with the original multi-scale features, and the edge positioning is more obvious and clear.
FIG. 3 shows a flow chart of a method of inspecting templates in an optimized disparity model of the present invention.
According to the embodiment of the invention, the difference model is trained by using the training image of the standard silicon wafer, the inspection templates in the difference model are registered, and the corresponding inspection templates are optimized, specifically:
S302, obtaining an original image of a standard silicon wafer as a training image of a difference model, locally dividing the training image from a template image of an inspection template in the difference model, carrying out local feature coding in a local area, and setting two learning training paths of rotation and translation in the difference model;
S304, coding single pixel points in a local area by utilizing a multi-layer sensing module to obtain local single point characteristics, carrying out average pooling on the local single point characteristics to obtain local integral characteristics of the local area, and splicing the local single point characteristics and the local integral characteristics in characteristic channel dimensions;
S306, activating the spliced features by using an activation function, performing batch standardization and maximum pooling operation, outputting final local features, and acquiring rotation and translation information of a training image and a template image according to the final local features in two learning training paths;
S308, obtaining attention weights according to the similarity between the training image and the local area of the template image in the rotation and translation information, weighting the local area of the training image by using the attention weights, and adding the weighted local area of the training image to the template image through residual errors;
S310, obtaining optimized template images in the rotation and translation two learning training paths, and optimizing an inspection template in the difference model.
It should be noted that, two learning training paths of rotation and translation are set in the difference model, according to single pixel point coding and according to spatial distribution feature coding of the whole local area, single pixel points of the local area are associated with the whole spatial distribution, rotation and translation information of the local feature training image and the template image is obtained, similarity of the local area in the training image and the template image is calculated to perform standardization processing, an attention weight matrix is generated, importance degrees of features of different areas are represented, and the local area more similar to the training image and the template image is more obvious in template optimization.
Obtaining an optimized inspection template in the difference model, and determining a gray value parameter of the difference model according to actual conditions, wherein the gray value parameter comprises a minimum value alpha of a gray value difference value between a current image and a template image, and the minimum value alpha is an absolute threshold; the minimum relative gray level difference beta between the current image and the template image and between the current image and the template image is a relative threshold value, and the pixel points which are too bright and too dark are aimed at; calculating a maximum gray threshold D h and a minimum gray threshold D l according to the gray value parameters, specifically:
Dh＝b(x,y)+max(α,βm(x,y))
Dl＝b(x,y)-max(α,βm(x,y))
Wherein b (x, y) represents a template image, and m (x, y) represents a difference image of the difference model;
And obtaining a maximum gray value image and a minimum gray value image corresponding to the silicon chip according to the maximum gray threshold value and the minimum gray threshold value, and storing the maximum gray value image and the minimum gray value image into the difference model.
The method comprises the steps of acquiring a silicon wafer image to be detected by using a silicon wafer image acquisition device, and performing automatic global threshold segmentation and rectangular segmentation on the silicon wafer image to be detected to acquire an interested region; introducing a region of interest into a difference model to extract edge characteristics of a silicon wafer, calculating similarity between an image to be detected and a detection template according to normalized product correlation by utilizing the edge characteristics, performing hierarchical search and matching in an image pyramid according to the similarity, and performing matching from the pyramid tip; searching surrounding local areas in the next layer of the image pyramid according to the matching position of the pyramid tip, continuing to match, and acquiring areas with similarity meeting preset standards from the corresponding search image of the silicon wafer image to be detected, so as to finish matching and positioning of the silicon wafer image to be detected; and carrying out affine transformation by using the homogeneous rotation matrix after matching and positioning, and carrying out position correction.
Obtaining an average gray value of a region of interest in a silicon wafer image to be detected, and scaling the average gray value of the region of interest according to an inspection template of a difference model; comparing the scaled average gray value with the maximum gray value image and the minimum gray value image in the difference model, taking the region which is larger than the gray value corresponding to the maximum gray value image and the region which is smaller than the gray value corresponding to the minimum gray value image as a silicon wafer defect region, and visually marking the exact position of the defect region.
Template matching can effectively detect known patterns or defects on a silicon wafer. The template matching provides accurate positioning of the characteristics or defects, accurately positions the matching positions of the template and the silicon wafer image, and is beneficial to identification and evaluation of the defects. Template matching may be robust to rotation and scaling when using appropriate techniques, where defects or features may appear in different directions or sizes. Template matching is less sensitive to changes in illumination conditions than other methods.
According to the embodiment of the invention, the defect detection result of the silicon wafer in the historical time step is obtained, the defect image is segmented according to the defect detection result, different clusters are obtained through clustering of the defect image, the main component characteristics of the defect characteristics in the clusters are extracted, the defect types of the different clusters are determined by utilizing the main component characteristics, and the defect types of the defective silicon wafer are identified by utilizing the cluster training classifier with the defect type label; counting occurrence frequencies of defect categories in historical time steps, and screening high-frequency defect categories by using a counting result; and (3) carrying out silicon wafer production tracing according to the high-frequency defect types, acquiring defect causes of the defect types by utilizing the big data, carrying out fine granularity comparison according to the defect causes to carry out defect judgment, roughly dividing the defect causes into raw material factors, production process factors and the like, acquiring and screening silicon wafer production procedures when the process factors are judged to be generated, acquiring deviation between production monitoring data and standard production data in the silicon wafer production procedures, carrying out defect tracing positioning according to the deviation, correcting production parameters, classifying the defective silicon wafers according to defect type labels and defect geometric characteristics, presetting threshold information, and collecting the defective silicon wafers conforming to the threshold information for secondary use.
FIG. 4 shows a block diagram of a system for detecting surface defects of a silicon wafer based on template matching according to the present invention.
The second aspect of the present invention also provides a system 4 for detecting surface defects of a silicon wafer based on template matching, which comprises: the memory 41 and the processor 42, wherein the memory comprises a silicon wafer surface defect detection method program based on template matching, and the silicon wafer surface defect detection method program based on template matching realizes the following steps when being executed by the processor:
acquiring good silicon wafer images, performing automatic global threshold segmentation and rectangular segmentation on the good silicon wafer images, removing silicon wafer background, and acquiring an interested region;
Performing mean filtering on the region of interest, performing pyramid operation on the filtered image, creating a corresponding shape outline, and constructing a difference model;
Training the difference model by using a training image of a standard silicon wafer, registering the inspection templates in the difference model, and optimizing the corresponding inspection templates;
the method comprises the steps of obtaining a silicon wafer image to be detected, extracting an interested region, guiding the interested region of the silicon wafer image to be detected into a difference model, utilizing an inspection template to match, extracting matching differences according to a matching result, and visually displaying the difference part.
The third aspect of the present invention also provides a computer readable storage medium, where the computer readable storage medium includes a method program for detecting a surface defect of a silicon wafer based on template matching, and when the method program is executed by a processor, the steps of the method for detecting a surface defect of a silicon wafer based on template matching are implemented.
In the several embodiments provided by the present application, it should be understood that the disclosed apparatus and method may be implemented in other ways. The above described device embodiments are only illustrative, e.g. the division of the units is only one logical function division, and there may be other divisions in practice, such as: multiple units or components may be combined or may be integrated into another system, or some features may be omitted, or not performed. In addition, the various components shown or discussed may be coupled or directly coupled or communicatively coupled to each other via some interface, whether indirectly coupled or communicatively coupled to devices or units, whether electrically, mechanically, or otherwise.
The units described above as separate components may or may not be physically separate, and components shown as units may or may not be physical units; can be located in one place or distributed to a plurality of network units; some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.
In addition, each functional unit in each embodiment of the present invention may be integrated in one processing unit, or each unit may be separately used as one unit, or two or more units may be integrated in one unit; the integrated units may be implemented in hardware or in hardware plus software functional units.
Those of ordinary skill in the art will appreciate that: all or part of the steps for implementing the above method embodiments may be implemented by hardware related to program instructions, and the foregoing program may be stored in a computer readable storage medium, where the program, when executed, performs steps including the above method embodiments; and the aforementioned storage medium includes: a mobile storage device, a Read-On Memory (ROM), a random access Memory (RAM, ran dom Access Memory), a magnetic disk or an optical disk, or the like, which can store program codes.
Or the above-described integrated units of the invention may be stored in a computer-readable storage medium if implemented in the form of software functional modules and sold or used as separate products. Based on such understanding, the technical solutions of the embodiments of the present invention may be embodied in essence or a part contributing to the prior art in the form of a software product stored in a storage medium, including several instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to execute all or part of the methods described in the embodiments of the present invention. And the aforementioned storage medium includes: a removable storage device, ROM, RAM, magnetic or optical disk, or other medium capable of storing program code.
The foregoing is merely illustrative of the present invention, and the present invention is not limited thereto, and any person skilled in the art will readily recognize that variations or substitutions are within the scope of the present invention. Therefore, the protection scope of the present invention shall be subject to the protection scope of the claims.
Claims (8)
1. A method for detecting surface defects of a silicon wafer based on template matching is characterized by comprising the following steps:
acquiring good silicon wafer images, performing automatic global threshold segmentation and rectangular segmentation on the good silicon wafer images, removing silicon wafer background, and acquiring an interested region;
Performing mean filtering on the region of interest, performing pyramid operation on the filtered image, creating a corresponding shape outline, and constructing a difference model;
Training the difference model by using a training image of a standard silicon wafer, registering the inspection templates in the difference model, and optimizing the corresponding inspection templates;
the method comprises the steps of obtaining a silicon wafer image to be detected, extracting an interested region, guiding the interested region of the silicon wafer image to be detected into a difference model, utilizing an inspection template to match, extracting matching differences according to a matching result, and visually displaying the difference part.
2. The method for detecting the surface defects of the silicon wafer based on the template matching according to claim 1, wherein the method is characterized in that the silicon wafer image with good quality is subjected to automatic global threshold segmentation and rectangular segmentation, and specifically comprises the following steps:
Acquiring good silicon wafer images, dividing the good silicon wafer images into a background area and a foreground area according to pixel intensity values, and automatically determining an optimal threshold value by adopting an OSTU algorithm;
carrying out automatic global threshold segmentation on the good silicon wafer image through the optimal threshold value, and acquiring a placement area of the good silicon wafer in a foreground area;
And obtaining edge characteristics in the placement area of the good silicon wafer by using a Canny edge detection algorithm, performing rectangular segmentation based on the edge characteristics, and further segmenting the background to obtain an interested area.
3. The method for detecting the surface defects of the silicon wafer based on template matching according to claim 1, wherein the method is characterized in that the region of interest is subjected to mean value filtering, the filtered image is subjected to pyramid operation, a corresponding shape outline is created, and a difference model is constructed, specifically:
The method comprises the steps of carrying out mean filtering on the region of interest, reducing high-frequency noise through the change of smooth pixel intensity, extracting original multi-scale features of the region of interest by utilizing an image pyramid, and embedding a channel attention mechanism to filter the original multi-scale features to obtain filtered multi-scale features;
Performing convolution operation on the multi-scale features to obtain salient features, obtaining salient objects with different sizes and shapes, and splicing the salient features with the same resolution in the original multi-scale features after feature up-sampling;
Fully fusing the spliced features by utilizing convolution operation, supplementing the space detail features of the salient objects, and creating the shape outline of the silicon wafer in the region of interest;
And constructing the difference model by taking the obtained shape outline as a template image of the difference model and taking an original image of a standard silicon wafer as input.
4. The method for detecting the surface defects of the silicon wafer based on the template matching according to claim 1, wherein the difference model is trained by using training images of a standard silicon wafer, the inspection templates in the difference model are registered, and the corresponding inspection templates are optimized, specifically:
obtaining an original image of a standard silicon wafer as a training image of a difference model, carrying out local division on the training image and a template image of an inspection template in the difference model, carrying out local feature coding in a local area, and setting two rotation and translation learning training paths in the difference model;
coding single pixel points in a local area by utilizing a multi-layer sensing module to obtain local single point characteristics, carrying out average pooling on the local single point characteristics to obtain local integral characteristics of the local area, and splicing the local single point characteristics and the local integral characteristics in characteristic channel dimensions;
activating the spliced features by using an activation function, performing batch standardization and maximum pooling operation, outputting final local features, and acquiring rotation and translation information of a training image and a template image according to the final local features in two learning training paths;
Obtaining attention weights according to the similarity between the training image and the local area of the template image in the rotation and translation information, weighting the local area of the training image by using the attention weights, and adding the weighted local area of the training image to the template image through residual errors;
And acquiring optimized template images in the rotation and translation two learning training paths, and optimizing an inspection template in the difference model.
5. The method for detecting the surface defects of the silicon wafer based on the template matching according to claim 4, wherein an optimized inspection template in a difference model is obtained, gray value parameters of the difference model are determined, and a maximum gray threshold value and a minimum gray threshold value are calculated according to the gray value parameters;
And obtaining a maximum gray value image and a minimum gray value image corresponding to the silicon chip according to the maximum gray threshold value and the minimum gray threshold value, and storing the maximum gray value image and the minimum gray value image into the difference model.
6. The method for detecting the surface defects of the silicon wafer based on the template matching according to claim 1, wherein the method is characterized in that the silicon wafer image to be detected is obtained to extract the region of interest, the region of interest of the silicon wafer image to be detected is led into a difference model, and the matching is carried out by using an inspection template, specifically:
acquiring a silicon wafer image to be detected by using a silicon wafer image acquisition device, and performing automatic global threshold segmentation and rectangular segmentation on the silicon wafer image to be detected to acquire an interested region;
Introducing a region of interest into a difference model to extract edge characteristics of a silicon wafer, calculating similarity between an image to be detected and a detection template according to normalized product correlation by utilizing the edge characteristics, performing hierarchical search and matching in an image pyramid according to the similarity, and performing matching from the pyramid tip;
searching surrounding local areas in the next layer of the image pyramid according to the matching position of the pyramid tip, continuing to match, and acquiring areas with similarity meeting preset standards from the corresponding search image of the silicon wafer image to be detected, so as to finish matching and positioning of the silicon wafer image to be detected;
and carrying out affine transformation by using the homogeneous rotation matrix after matching and positioning, and carrying out position correction.
7. The method for detecting the surface defects of the silicon wafer based on the template matching according to claim 1, wherein the matching difference is extracted according to the matching result, and the difference part is visually displayed, specifically:
obtaining an average gray value of a region of interest in a silicon wafer image to be detected, and scaling the average gray value of the region of interest according to an inspection template of a difference model;
And comparing the scaled average gray value with the maximum gray value image and the minimum gray value image in the difference model, outputting a silicon wafer defect area according to a comparison result, and visually marking the exact position of the defect area.
8. A silicon wafer surface defect detection system based on template matching is characterized in that the system comprises: the device comprises a memory and a processor, wherein the memory comprises a silicon wafer surface defect detection method program based on template matching, and the silicon wafer surface defect detection method program based on the template matching realizes the following steps when being executed by the processor:
acquiring good silicon wafer images, performing automatic global threshold segmentation and rectangular segmentation on the good silicon wafer images, removing silicon wafer background, and acquiring an interested region;
Performing mean filtering on the region of interest, performing pyramid operation on the filtered image, creating a corresponding shape outline, and constructing a difference model;
Training the difference model by using a training image of a standard silicon wafer, registering the inspection templates in the difference model, and optimizing the corresponding inspection templates;
the method comprises the steps of obtaining a silicon wafer image to be detected, extracting an interested region, guiding the interested region of the silicon wafer image to be detected into a difference model, utilizing an inspection template to match, extracting matching differences according to a matching result, and visually displaying the difference part.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202410139542.0A CN117974601A (en) | 2024-02-01 | 2024-02-01 | Method and system for detecting surface defects of silicon wafer based on template matching |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202410139542.0A CN117974601A (en) | 2024-02-01 | 2024-02-01 | Method and system for detecting surface defects of silicon wafer based on template matching |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN117974601A true CN117974601A (en) | 2024-05-03 |
Family
ID=90861147
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202410139542.0A Pending CN117974601A (en) | 2024-02-01 | 2024-02-01 | Method and system for detecting surface defects of silicon wafer based on template matching |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN117974601A (en) |
-
2024
- 2024-02-01 CN CN202410139542.0A patent/CN117974601A/en active Pending
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109978839B (en) | Method for detecting wafer low-texture defects | |
| KR101934313B1 (en) | System, method and computer program product for detection of defects within inspection images | |
| US20130202188A1 (en) | Defect inspection method, defect inspection apparatus, program product and output unit | |
| CN113724231B (en) | Industrial defect detection method based on semantic segmentation and target detection fusion model | |
| CN110378313B (en) | Cell cluster identification method and device and electronic equipment | |
| CN110930390B (en) | Chip pin missing detection method based on semi-supervised deep learning | |
| CN115861291B (en) | Chip circuit board production defect detection method based on machine vision | |
| CN107966454A (en) | A kind of end plug defect detecting device and detection method based on FPGA | |
| JP2000137804A (en) | Method and system for abnormality detection of digital image and storage medium for same | |
| CN111415329A (en) | Workpiece surface defect detection method based on deep learning | |
| CN114820625B (en) | Automobile top block defect detection method | |
| CN110648330B (en) | Defect detection method for camera glass | |
| CN113221881B (en) | Multi-level smart phone screen defect detection method | |
| CN113096119A (en) | Method and device for classifying wafer defects, electronic equipment and storage medium | |
| CN108665464A (en) | A kind of foreign matter detecting method based on morphologic high tension electric tower and high-tension bus-bar | |
| CN113240623B (en) | Pavement disease detection method and device | |
| CN116485779B (en) | Adaptive wafer defect detection method and device, electronic equipment and storage medium | |
| CN115797813B (en) | Water environment pollution detection method based on aerial image | |
| CN115290663A (en) | Mini LED wafer appearance defect detection method based on optical detection | |
| CN111695373A (en) | Zebra crossing positioning method, system, medium and device | |
| CN115830004A (en) | Surface defect detection method, device, computer equipment and storage medium | |
| CN115082776A (en) | Electric energy meter automatic detection system and method based on image recognition | |
| CN115797314B (en) | Method, system, equipment and storage medium for detecting surface defects of parts | |
| CN116433978A (en) | Automatic generation and automatic labeling method and device for high-quality flaw image | |
| CN113192018B (en) | Water-cooled wall surface defect video identification method based on fast segmentation convolutional neural network |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication |