CN116228780A - Silicon wafer defect detection method and system based on computer vision - Google Patents
Silicon wafer defect detection method and system based on computer vision Download PDFInfo
- Publication number
- CN116228780A CN116228780A CN202310521596.9A CN202310521596A CN116228780A CN 116228780 A CN116228780 A CN 116228780A CN 202310521596 A CN202310521596 A CN 202310521596A CN 116228780 A CN116228780 A CN 116228780A
- Authority
- CN
- China
- Prior art keywords
- preset
- silicon wafer
- defect
- candidate frame
- pixel
- 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.)
- Granted
Links
- 230000007547 defect Effects 0.000 title claims abstract description 147
- XUIMIQQOPSSXEZ-UHFFFAOYSA-N Silicon Chemical compound [Si] XUIMIQQOPSSXEZ-UHFFFAOYSA-N 0.000 title claims abstract description 131
- 229910052710 silicon Inorganic materials 0.000 title claims abstract description 131
- 239000010703 silicon Substances 0.000 title claims abstract description 131
- 238000001514 detection method Methods 0.000 title claims abstract description 47
- 230000002146 bilateral effect Effects 0.000 claims abstract description 63
- 230000009467 reduction Effects 0.000 claims abstract description 51
- 239000013598 vector Substances 0.000 claims abstract description 42
- 238000000034 method Methods 0.000 claims abstract description 29
- 238000005457 optimization Methods 0.000 claims abstract description 29
- 238000013507 mapping Methods 0.000 claims abstract description 20
- 238000004458 analytical method Methods 0.000 claims abstract description 13
- 238000001914 filtration Methods 0.000 claims description 48
- 238000004422 calculation algorithm Methods 0.000 claims description 45
- 230000011218 segmentation Effects 0.000 claims description 30
- 239000011159 matrix material Substances 0.000 claims description 22
- 238000009826 distribution Methods 0.000 claims description 19
- 230000002950 deficient Effects 0.000 claims description 12
- 230000001629 suppression Effects 0.000 claims description 10
- 238000012545 processing Methods 0.000 claims description 7
- 230000015556 catabolic process Effects 0.000 claims description 3
- 238000006731 degradation reaction Methods 0.000 claims description 3
- 238000010606 normalization Methods 0.000 claims description 3
- 238000001228 spectrum Methods 0.000 claims description 3
- 238000010586 diagram Methods 0.000 abstract description 8
- 238000013473 artificial intelligence Methods 0.000 abstract description 7
- 238000005516 engineering process Methods 0.000 abstract description 3
- 230000006870 function Effects 0.000 description 35
- 238000003384 imaging method Methods 0.000 description 6
- 230000000694 effects Effects 0.000 description 5
- 238000003709 image segmentation Methods 0.000 description 5
- 238000005259 measurement Methods 0.000 description 3
- 238000013135 deep learning Methods 0.000 description 2
- 230000000737 periodic effect Effects 0.000 description 2
- 230000008569 process Effects 0.000 description 2
- 150000003839 salts Chemical class 0.000 description 2
- 239000004065 semiconductor Substances 0.000 description 2
- 235000002566 Capsicum Nutrition 0.000 description 1
- 206010063385 Intellectualisation Diseases 0.000 description 1
- 239000006002 Pepper Substances 0.000 description 1
- 235000016761 Piper aduncum Nutrition 0.000 description 1
- 235000017804 Piper guineense Nutrition 0.000 description 1
- 244000203593 Piper nigrum Species 0.000 description 1
- 235000008184 Piper nigrum Nutrition 0.000 description 1
- 230000004913 activation Effects 0.000 description 1
- 208000003464 asthenopia Diseases 0.000 description 1
- 230000005540 biological transmission Effects 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 238000004891 communication Methods 0.000 description 1
- 238000004590 computer program Methods 0.000 description 1
- 238000013500 data storage Methods 0.000 description 1
- 230000001419 dependent effect Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000018109 developmental process Effects 0.000 description 1
- 238000002059 diagnostic imaging Methods 0.000 description 1
- 238000009499 grossing Methods 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 230000007246 mechanism Effects 0.000 description 1
- 238000000386 microscopy Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000003062 neural network model Methods 0.000 description 1
- 235000012434 pretzels Nutrition 0.000 description 1
- 238000012163 sequencing technique Methods 0.000 description 1
- 241000894007 species Species 0.000 description 1
- 238000003860 storage Methods 0.000 description 1
- 238000012360 testing method Methods 0.000 description 1
- 230000009466 transformation Effects 0.000 description 1
- 230000001131 transforming effect Effects 0.000 description 1
- 238000013519 translation Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/0002—Inspection of images, e.g. flaw detection
- G06T7/0004—Industrial image inspection
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T5/00—Image enhancement or restoration
- G06T5/20—Image enhancement or restoration by the use of local operators
-
- G06T5/70—
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/11—Region-based segmentation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10024—Color image
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20024—Filtering details
- G06T2207/20028—Bilateral filtering
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30108—Industrial image inspection
Abstract
The invention relates to artificial intelligence technology, and discloses a silicon wafer defect detection method and system based on computer vision, wherein the method comprises the following steps: dividing a silicon wafer picture to obtain a defect division diagram, and carrying out bilateral noise reduction and wiener noise reduction on the defect division diagram to obtain a noise reduction optimization diagram; generating a primary candidate frame, extracting candidate features of the primary candidate frame, and mapping the candidate features to a preset probability space to obtain defect types and defect probability values; selecting a target defect, and mapping a primary candidate frame corresponding to the candidate feature to a vector space to obtain an offset vector; and performing position adjustment on the primary candidate frame to obtain standard position information of the primary candidate frame, optimizing the adjusted primary candidate frame to obtain a standard candidate frame, and taking the standard candidate frame, the corresponding standard position information and the target defect as defect analysis results of the silicon wafer. The invention can improve the accuracy of silicon wafer defect detection.
Description
Technical Field
The invention relates to the technical field of artificial intelligence, in particular to a silicon wafer defect detection method and system based on computer vision.
Background
With the continuous development of the semiconductor industry, important components in the semiconductor industry are detected when the defects of the silicon wafer are detected. In order to improve the efficiency and accuracy of silicon wafer defect detection, a machine vision is used for analyzing a silicon wafer image, so that the accurate detection of the silicon wafer defect is realized.
The traditional silicon wafer defect detection needs a large amount of manual operation, consumes a large amount of time and energy, and cannot meet the high-efficiency requirements of modern production. In practical application, the traditional silicon wafer defect detection is unstable due to interference of human factors such as visual fatigue, subjectivity and the like, and the precision and reliability of the silicon wafer defect detection result are difficult to ensure.
Disclosure of Invention
The invention provides a silicon wafer defect detection method and system based on computer vision, and mainly aims to solve the problem of lower accuracy in silicon wafer defect detection.
In order to achieve the above object, the present invention provides a method for detecting a defect of a silicon wafer based on computer vision, comprising:
carrying out camera calibration on preset microscopic equipment to obtain calibration equipment, and acquiring a silicon wafer picture of a preset silicon wafer by using the calibration equipment;
dividing the silicon wafer picture by using a preset Markov random field division algorithm to obtain a defect division picture, carrying out bilateral noise reduction on the defect division picture by using a preset bilateral filtering function to obtain a bilateral filtering picture, and carrying out wiener noise reduction on the bilateral filtering picture by using a preset wiener filter to obtain a noise reduction optimization picture;
Generating a primary candidate frame according to the noise reduction optimization graph by using a preset area proposal network, extracting candidate features of the primary candidate frame, and mapping the candidate features to a preset probability space by using a preset classification sub-network to obtain defect categories and defect probability values corresponding to the candidate features;
selecting a defect category with the defect probability value larger than a preset probability threshold as a target defect of the candidate feature, and mapping a primary candidate frame corresponding to the candidate feature to a preset vector space according to a preset regression sub-network to obtain an offset vector corresponding to the primary candidate frame;
and carrying out position adjustment on the primary candidate frame according to the offset vector to obtain standard position information of the primary candidate frame, optimizing the adjusted primary candidate frame by using a preset non-maximum suppression algorithm to obtain a standard candidate frame, and taking the standard candidate frame, the corresponding standard position information and the target defect as defect analysis results of the silicon wafer.
Optionally, the calibrating the camera of the preset microscopic device to obtain the calibrating device includes:
acquiring a preset black-and-white chessboard graph as a calibration plate of the microscopic equipment;
Acquiring a calibration image of the calibration plate under a preset shooting condition by using a camera in the microscopic equipment;
detecting and positioning the calibration image by using a corner detection algorithm to obtain the position information of preset feature points;
according to preset calibration plate parameters, combining the position information, and calculating an internal parameter matrix of the camera by using a camera calibration algorithm;
according to the position information and the shooting conditions, calculating an external parameter matrix of the camera by using a preset camera pose estimation algorithm;
integrating the internal parameter matrix and the external parameter matrix into a camera parameter matrix, and determining microscopic equipment capable of correcting the shot picture by using the camera parameter matrix as calibration equipment.
Optionally, the segmenting the silicon wafer picture by using a preset markov random field segmentation algorithm to obtain a defect segmentation map includes:
acquiring the region category of a pixel of a preset silicon wafer picture, and defining a potential function of the silicon wafer picture according to the region category;
calculating pixel region distribution of the silicon wafer picture according to the potential function;
the maximum pixel region distribution is determined as a defect segmentation map.
Optionally, the defining the potential function of the silicon wafer picture according to the region category includes:
defining a potential function of the silicon wafer picture by using the following definition formula:
wherein ,
as a potential function +.>
Indicate->
Regional category to which each pixel belongs,>
indicate->
Regional category to which each pixel belongs,>
representing two unequal pixel sequence numbers, < ->
Indicate->
The individual pixels belong to->
Probability of->
Indicate->
And->
The individual pixels belong to->
and />
Is (are) joint probability->
Is the total number of pixel sequence numbers, +.>
Is a preset parameter for controlling smoothness.
Optionally, the obtaining the region category to which the pixel of the preset silicon wafer picture belongs includes:
performing pixel coding on the silicon wafer picture to obtain a silicon wafer pixel;
dividing the silicon wafer pixels into preset pixel areas;
and classifying the pixel areas by using a preset pixel label to obtain area categories.
Optionally, the calculating the distribution of the pixel areas of the silicon wafer picture according to the potential function includes:
acquiring a pixel gray value of the silicon wafer picture;
calculating the distribution of pixel areas of the silicon wafer picture according to the pixel gray values by using the following probability distribution formula:
wherein ,
Indicate->
Pixel gray value of individual pixels +.>
Representing the distribution of pixel areas of a silicon wafer picture, < >>
Indicate->
Regional category to which each pixel belongs,>
as a potential function +.>
Is a preset normalization constant.
Optionally, the performing bilateral noise reduction on the defect segmentation graph by using a preset bilateral filtering function to obtain a bilateral filtering graph includes:
obtaining a filtered pixel value corresponding to a pixel in the defect segmentation map by using the following bilateral filtering formula:
wherein ,
representing filtered pixel points subjected to bilateral filtering processing>
The corresponding filtered pixel value is used to determine,
defective pixel point +.>
Corresponding defective pixel value,/->
Representing preset filtering pixel points
And defective pixel->
Spatial distance weight between->
Representing filtered pixel points +.>
And defective pixel->
Preset gray value similarity weight, +.>
Representing filtered pixel points +.>
Neighborhood of->
Is the weight sum of all gray value similarity weights;
and collecting the filtered pixel values into a double-sided filter map.
Optionally, the wiener denoising the bilateral filtering graph by using a preset wiener filter to obtain a denoising optimization graph, which includes:
and carrying out wiener denoising on the bilateral filter graph by using a preset wiener filter by using the following wiener filter formula:
wherein ,
for the filtered noise reduction optimization map, +.>
For bilateral filter patterns, < >>
Is a preset band-pass filter +.>
Fourier transform of a predetermined degradation function, +.>
Signal power spectrum for bilateral filter patterns, +.>
Is the preset noise power.
Optionally, the mapping, according to a preset regression sub-network, the primary candidate frame corresponding to the candidate feature to a preset vector space to obtain an offset vector corresponding to the primary candidate frame includes:
calculating an offset vector corresponding to the primary candidate frame by utilizing a boundary regression formula in the regression sub-network:
wherein ,
an abscissa representing the central coordinates of said primary candidate frame,/->
An ordinate representing the center coordinates of the primary candidate frame,/->
An abscissa representing the center coordinates of the reference frame corresponding to the primary candidate frame, +.>
Showing the ordinate of the center coordinates of the reference frame corresponding to the primary candidate frame, +.>
Representing the width of the primary candidate box, +.>
Representing the height, +.>
Representing the width of the reference frame corresponding to the primary candidate frame,/or->
Representing the height of the reference frame corresponding to the primary candidate frame,/>
For the abscissa position of the primary candidate frame relative to the corresponding reference frame, >
For the ordinate position of the primary candidate frame relative to the corresponding reference frame,>
for the width of the primary candidate frame relative to the corresponding reference frame,/a>
For the offset height of the primary candidate frame relative to the corresponding reference frame,/>
Is an offset vector.
In order to solve the above problems, the present invention further provides a silicon wafer defect detection system based on computer vision, the system comprising:
and a picture acquisition module: carrying out camera calibration on preset microscopic equipment to obtain calibration equipment, and acquiring a silicon wafer picture of a preset silicon wafer by using the calibration equipment;
noise reduction optimization module: dividing the silicon wafer picture by using a preset Markov random field division algorithm to obtain a defect division picture, carrying out bilateral noise reduction on the defect division picture by using a preset bilateral filtering function to obtain a bilateral filtering picture, and carrying out wiener noise reduction on the bilateral filtering picture by using a preset wiener filter to obtain a noise reduction optimization picture;
determining a candidate frame module: generating a primary candidate frame according to the noise reduction optimization graph by using a preset area proposal network, extracting candidate features of the primary candidate frame, and mapping the candidate features to a preset probability space by using a preset classification sub-network to obtain defect categories and defect probability values corresponding to the candidate features;
And (3) adjusting a candidate frame module: selecting a defect category with the defect probability value larger than a preset probability threshold as a target defect of the candidate feature, and mapping a primary candidate frame corresponding to the candidate feature to a preset vector space according to a preset regression sub-network to obtain an offset vector corresponding to the primary candidate frame;
and a result analysis module: and carrying out position adjustment on the primary candidate frame according to the offset vector to obtain standard position information of the primary candidate frame, optimizing the adjusted primary candidate frame by using a preset non-maximum suppression algorithm to obtain a standard candidate frame, and taking the standard candidate frame, the corresponding standard position information and the target defect as defect analysis results of the silicon wafer.
According to the embodiment of the invention, the silicon wafer picture is segmented by utilizing a preset Markov random field segmentation algorithm to obtain a defect segmentation map, a preset bilateral filtering function and a wiener filter are utilized for double filtering to obtain a noise reduction optimization map, and the bilateral filtering and the wiener filter are combined for use, so that the advantages of the two algorithms can be fully exerted, and a better noise reduction effect is achieved; generating a primary candidate frame according to the noise reduction optimization graph by using a preset area proposal network, extracting candidate features of the primary candidate frame, and mapping the candidate features to a preset probability space by using a preset classification sub-network to obtain defect categories and defect probability values corresponding to the candidate features; selecting a defect category with the defect probability value larger than a preset probability threshold as a target defect of the candidate feature, mapping a primary candidate frame corresponding to the candidate feature to a preset vector space according to a preset regression sub-network to obtain an offset vector corresponding to the primary candidate frame, determining the offset vector of the candidate frame by using the regression sub-network, further improving the accuracy of silicon wafer defect detection, reducing the false detection rate and improving the detection rate by obtaining a more accurate candidate frame position, and having higher application value in actual scenes; and carrying out position adjustment on the primary candidate frame according to the offset vector to obtain standard position information of the primary candidate frame, optimizing the adjusted primary candidate frame by using a preset non-maximum suppression algorithm to obtain a standard candidate frame, taking the standard candidate frame, the corresponding standard position information and the target defect as defect analysis results of the silicon wafer, and filtering the overlapped candidate frames by using the non-maximum suppression algorithm to improve the accuracy and speed of target detection. Therefore, the silicon wafer defect detection method and system based on computer vision can solve the problem of lower silicon wafer defect detection accuracy.
Drawings
FIG. 1 is a schematic flow chart of a method for detecting defects of a silicon wafer based on computer vision according to an embodiment of the invention;
FIG. 2 is a flowchart of a method for obtaining a defect segmentation map according to an embodiment of the present invention;
FIG. 3 is a flow chart of acquiring region categories according to an embodiment of the present invention;
FIG. 4 is a functional block diagram of a silicon wafer defect detection system based on computer vision according to an embodiment of the present invention;
the achievement of the objects, functional features and advantages of the present invention will be further described with reference to the accompanying drawings, in conjunction with the embodiments.
Detailed Description
It should be understood that the specific embodiments described herein are for purposes of illustration only and are not intended to limit the scope of the invention.
The embodiment of the application provides a silicon wafer defect detection method based on computer vision. The execution subject of the method for detecting the defects of the silicon wafer based on the computer vision comprises at least one of a server, a terminal and the like which can be configured to execute the method provided by the embodiment of the application. In other words, the method for detecting the defect of the silicon chip based on the computer vision can be executed by software or hardware installed in a terminal device or a server device, wherein the software can be a blockchain platform. The service end includes but is not limited to: a single server, a server cluster, a cloud server or a cloud server cluster, and the like. The server may be an independent server, or may be a cloud server that provides cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, content delivery networks (ContentDelivery Network, CDN), and basic cloud computing services such as big data and artificial intelligence platforms.
Referring to fig. 1, a flow chart of a method for detecting defects of a silicon wafer based on computer vision according to an embodiment of the invention is shown. In this embodiment, the method for detecting a silicon wafer defect based on computer vision includes:
s1, calibrating a camera of preset microscopic equipment to obtain calibration equipment, and acquiring a silicon wafer picture of a preset silicon wafer by using the calibration equipment;
because of lens distortion of microscopic equipment, focusing at different positions and the like, distortion phenomenon can exist in directly obtained silicon wafer pictures, and accuracy of image processing results is affected. Therefore, the microscopic equipment needs to be calibrated by a camera, so that the distortion phenomenon is corrected, and a more accurate image processing result is obtained.
Furthermore, the camera calibration of the preset microscopic equipment can eliminate image distortion and distortion, so that the accuracy and the accuracy of pictures are improved, and in computer vision, the camera calibration can provide more accurate input data for tasks such as target tracking, target identification, image segmentation and the like. At the same time, camera calibration can help prevent fraud and erroneous measurements, which can result in erroneous size, shape, and position measurements due to image distortion if the microscopy device is not camera calibrated, increasing errors and uncertainty in the data.
In the embodiment of the present invention, the camera calibration is performed on a preset microscopic device to obtain a calibration device, including:
acquiring a preset black-and-white chessboard graph as a calibration plate of the microscopic equipment;
acquiring a calibration image of the calibration plate under a preset shooting condition by using a camera in the microscopic equipment;
detecting and positioning the calibration image by using a corner detection algorithm to obtain the position information of preset feature points;
according to preset calibration plate parameters, combining the position information, and calculating an internal parameter matrix of the camera by using a camera calibration algorithm;
according to the position information and the shooting conditions, calculating an external parameter matrix of the camera by using a preset camera pose estimation algorithm;
integrating the internal parameter matrix and the external parameter matrix into a camera parameter matrix, and determining microscopic equipment capable of correcting the shot picture by using the camera parameter matrix as calibration equipment.
In particular, the calibration plate is a special planar object, usually composed of a checkerboard or a lattice of dots with alternating black and white, capable of providing a known, regular object model as a reference for camera calibration.
Further, the calibration image is detected and positioned by using a corner detection algorithm, wherein the corner detection algorithm detects and positions characteristic points by detecting obvious corners (also called key points) in the image. Common corner detection algorithms include Harris corner detection, shi-Tomasi corner detection and the like.
In addition, the preset feature points are the corner points determined according to a corner point detection algorithm, and the sub-pixel level accurate positioning is carried out on the screened corner points, so that the position information of the feature points is obtained.
In detail, the internal parameter matrix refers to internal parameters in the imaging process of the camera, and generally includes focal length, principal point, pixel size, distortion parameters, and the like. The external parameter matrix refers to the spatial position and attitude information of the camera, and generally comprises a rotation matrix, a translation vector and the like. These parameters can be used to correct the lens distortion of the camera, correct the deformation of the image shot by the camera, and realize accurate measurement and positioning.
Additionally, the camera calibration algorithm refers to obtaining internal parameters (such as focal length, distortion, etc.) of the camera by performing a series of tests and calculations on the camera, and the camera calibration algorithm includes, but is not limited to, zhang's calibration algorithm, bouguet's calibration algorithm, tsai's calibration algorithm, etc.
Further, the camera pose estimation algorithm is to detect and track a specific object in a known scene, so as to determine the pose of the camera under the world coordinate system.
In the embodiment of the present invention, the obtaining, by using the calibration device, a silicon wafer picture of a preset silicon wafer includes:
connecting a camera in the calibration equipment with preset computer equipment;
adjusting parameters in the camera through the computer equipment until the imaging of the silicon chip in the view-finding frame of the camera reaches the preset light and shadow requirement;
and correcting the imaging of the silicon wafer by using a camera parameter matrix of the calibration equipment to obtain a silicon wafer picture.
In detail, the calibration equipment is used for acquiring the silicon wafer picture of the preset silicon wafer, so that the imaging quality of a camera can be greatly improved, the imaging accuracy and consistency of the camera are ensured, and the accuracy of the defect detection of the silicon wafer is improved. Meanwhile, the method has better stability and consistency, and the imaging is carried out by using cameras with the same parameters, so that the minimization of the difference between images can be ensured, and the repeatability and stability of the defect detection of the silicon wafer are further improved.
S2, segmenting the silicon wafer picture by using a preset Markov random field segmentation algorithm to obtain a defect segmentation map, carrying out bilateral noise reduction on the defect segmentation map by using a preset bilateral filtering function to obtain a bilateral filtering map, and carrying out wiener noise reduction on the bilateral filtering map by using a preset wiener filter to obtain a noise reduction optimization map;
In the embodiment of the invention, the Markov random field segmentation algorithm is an image segmentation method based on a probability map model, which converts an image segmentation problem into solving the maximum posterior probability on the probability map, models the image by using the Markov random field, and adds priori knowledge and constraint conditions in a segmentation result so as to improve the accuracy and stability of segmentation.
In detail, the preset Markov random field segmentation algorithm is utilized to segment the silicon wafer picture, so that the accuracy and stability of image segmentation are improved, the problems of edge blurring, missing segmentation, wrong segmentation and the like are avoided, various complex scenes such as image noise, complex textures, object overlapping and the like can be dealt with, the method has stronger robustness, and the method has wide application prospect, and can help to realize automation and intellectualization in the fields of medical imaging, target detection, computer vision, artificial intelligence and the like.
In the embodiment of the present invention, referring to fig. 2, the method for dividing the silicon wafer picture by using a preset markov random field dividing algorithm to obtain a defect dividing picture includes:
s21, obtaining the region category of a pixel of a preset silicon wafer picture, and defining a potential function of the silicon wafer picture according to the region category;
S22, calculating pixel region distribution of the silicon wafer picture according to the potential function;
s23, determining the maximum pixel area distribution as a defect segmentation map.
In detail, the defining the potential function of the silicon wafer picture according to the region category includes:
defining a potential function of the silicon wafer picture by using the following definition formula:
wherein ,
as a potential function +.>
Indicate->
Regional category to which each pixel belongs,>
indicate->
Regional category to which each pixel belongs,>
representing two unequal pixel sequence numbers, < ->
Indicate->
The individual pixels belong to->
Probability of->
Indicate->
And->
The individual pixels belong to->
and />
Is (are) joint probability->
Is the total number of pixel sequence numbers, +.>
Is a preset parameter for controlling smoothness.
Additionally, referring to fig. 3, the obtaining the region category to which the pixel of the preset silicon wafer picture belongs includes:
s31, carrying out pixel coding on the silicon wafer picture to obtain a silicon wafer pixel;
s32, dividing the silicon wafer pixels into preset pixel areas;
s33, classifying the pixel areas by using preset pixel labels to obtain area categories.
Further, the preset pixel label may be other characteristics such as brightness and color of the pixels in the silicon wafer picture. The silicon pixels are divided into a plurality of adjacent regions, each region containing a set of similar pixels, and each pixel is classified into a region class.
In detail, the potential function of the silicon wafer picture is defined according to the region category, wherein the potential function is also called an energy function and is used for measuring the rationality of each pixel or pixel combination in the image segmentation result, and the potential function plays a role in evaluating the segmentation result and is also a key of a Markov random field segmentation algorithm.
Specifically, the calculating the distribution of the pixel areas of the silicon wafer picture according to the potential function includes:
acquiring a pixel gray value of the silicon wafer picture;
calculating the distribution of pixel areas of the silicon wafer picture according to the pixel gray values by using the following probability distribution formula:
wherein ,
indicate->
Pixel gray value of individual pixels +.>
Representing the distribution of pixel areas of a silicon wafer picture, < >>
Indicate->
Regional category to which each pixel belongs,>
as a potential function +.>
Is a preset normalization constant.
In the embodiment of the present invention, the performing bilateral noise reduction on the defect segmentation graph by using a preset bilateral filtering function to obtain a bilateral filtering graph includes:
obtaining a filtered pixel value corresponding to a pixel in the defect segmentation map by using the following bilateral filtering formula:
wherein ,
representing filtered pixel points subjected to bilateral filtering processing >
The corresponding filtered pixel value is used to determine,
defective pixel point +.>
Corresponding defective pixel value,/->
Representing preset filtering pixel points
And defective pixel->
Spatial distance weight between->
Representing filtered pixel points +.>
And defective pixel->
Preset gray value similarity weight, +.>
Representing filtered pixel points +.>
Neighborhood of->
Is the weight sum of all gray value similarity weights;
and collecting the filtered pixel values into a double-sided filter map.
In detail, the bilateral noise reduction of the defect segmentation map by using the preset bilateral filtering function is a common noise reduction method in image processing, which can effectively remove noise in an image and retain edge information of the image. For the defect segmentation map, the bilateral filtering can remove noise in the image, so that a segmentation result is clearer and more accurate, and in addition, common noise types such as salt and pepper noise, gaussian noise and the like in the image can be effectively eliminated by the bilateral filtering function, so that the defect segmentation precision is improved.
Since bilateral filtering is mainly used to remove some non-deep learning noise such as gaussian noise and pretzel noise, other types of noise may exist, such as periodic noise, signal dependent noise, etc., and these noises may affect the effect of bilateral filtering. The wiener filter is mainly used for removing deep learning noise such as signal related noise, periodic noise and the like, so that the advantages of the two algorithms can be fully exerted by combining the bilateral filtering with the wiener filter, and a better noise reduction effect is achieved.
In the embodiment of the present invention, the wiener noise reduction is performed on the bilateral filtering graph by using a preset wiener filter to obtain a noise reduction optimization graph, which includes:
and carrying out wiener denoising on the bilateral filter graph by using a preset wiener filter by using the following wiener filter formula:
wherein ,
for the filtered noise reduction optimization map, +.>
For bilateral filter patterns, < >>
Is a preset band-pass filter +.>
Fourier transform of a predetermined degradation function, +.>
Signal power spectrum for bilateral filter patterns, +.>
Is the preset noise power.
Further, the bilateral filtering chart is subjected to wiener denoising by using a preset wiener filter, the defect segmentation chart is actually filtered by using a double filtering mode of combining bilateral filtering and wiener filtering, damage of smoothing processing on edge information can be reduced while image detail information is kept, the condition of blurring or blurring is avoided, airspace and gray scale of an image can be smoothed at the same time, and a good denoising effect is achieved on some complicated image noises (such as spiced salt noises and Gaussian noises).
S3, generating a primary candidate frame according to the noise reduction optimization graph by using a preset area proposal network, extracting candidate features of the primary candidate frame, and mapping the candidate features to a preset probability space by using a preset classification sub-network to obtain defect categories and defect probability values corresponding to the candidate features;
In an embodiment of the invention, the region proposal network (Region Proposal Network, RPN) is a neural network model capable of generating candidate bounding boxes on images, which is typically used as part of a target detection model. The RPN adopts an anchor-based idea, anchors with different length-width ratios and scales are placed on a bottom convolution feature diagram, and each anchor is processed through regression and classification of two branches, so that a series of candidate frames are obtained.
In an embodiment of the present invention, the generating the primary candidate frame according to the noise reduction optimization graph by using a preset area proposal network, for example, the image size of the noise reduction optimization graph is
And have->
Seed size and->
A ratio of species. For each pixel position on the noise reduction optimization graph +.>
Generating +.>
A reference frame which is flattened to a length of +.>
As output of the area proposal network, i.e. the primary candidate box.
In detail, the extracting the candidate features of the primary candidate frame includes:
mapping the candidate frames to a preset feature map;
dividing the primary candidate frames into a preset number of sub-selection frames;
Calculating the average value of the sub-selection frames one by one in the characteristic diagram;
and splicing all average values into average vectors with fixed lengths, and determining the average vectors as candidate features of the primary candidate frames.
Specifically, the preset feature map is an output of a layer in the sub-network of the area proposal network, and in this specific layer, each pixel point corresponds to a relatively small area in the original input image. For each candidate box, its corresponding position and size on the feature map can be calculated by transforming its coordinates. By using the feature map, feature vectors of sub-frames corresponding to the candidate frames can be extracted.
In the embodiment of the present invention, the candidate features are mapped to a preset probability space by using a preset classification sub-network, where the classification sub-network is generally composed of several fully connected layers, each layer uses an activation function (such as ReLU) to perform nonlinear transformation, and finally outputs a probability distribution, and the probability space refers to a probability value of each candidate feature classified into each defect class, where the probability value indicates a probability that the candidate feature belongs to the corresponding defect class, and this probability space is generally implemented by using a multi-element classifier (such as softmax classifier).
S4, selecting a defect type with the defect probability value larger than a preset probability threshold as a target defect of the candidate feature, and mapping a primary candidate frame corresponding to the candidate feature to a preset vector space according to a preset regression sub-network to obtain an offset vector corresponding to the primary candidate frame;
in the embodiment of the invention, the defect category with the defect probability value larger than the preset probability threshold is selected as the target defect of the candidate feature, so that the accuracy of the target defect can be ensured, and the accuracy of detecting the defect of the silicon wafer can be improved.
In the embodiment of the present invention, the mapping, according to a preset regression sub-network, the primary candidate frame corresponding to the candidate feature to a preset vector space to obtain an offset vector corresponding to the primary candidate frame includes:
calculating an offset vector corresponding to the primary candidate frame by utilizing a boundary regression formula in the regression sub-network:
wherein ,
an abscissa representing the central coordinates of said primary candidate frame,/->
An ordinate representing the center coordinates of the primary candidate frame,/->
An abscissa representing the center coordinates of the reference frame corresponding to the primary candidate frame, +.>
Showing the ordinate of the center coordinates of the reference frame corresponding to the primary candidate frame, +. >
Representing the width of the primary candidate box, +.>
Representing the height, +.>
Representing the width of the reference frame corresponding to the primary candidate frame,/or->
Representing the reference frame corresponding to the primary candidate frameHeight of->
For the abscissa position of the primary candidate frame relative to the corresponding reference frame,>
for the ordinate position of the primary candidate frame relative to the corresponding reference frame,>
for the width of the primary candidate frame relative to the corresponding reference frame,/a>
For the offset height of the primary candidate frame relative to the corresponding reference frame,/>
Is an offset vector.
In detail, the regression sub-network is a module commonly used in target detection for predicting the offset between a candidate frame and a real target frame, typically as part of a target detection model, for fine-tuning the position and size of the candidate frame to better match the target. The accuracy of the silicon wafer defect detection can be further improved by determining the offset vector of the candidate frame by using the regression sub-network. By obtaining more accurate candidate frame positions, the false detection rate can be reduced, the detection rate can be improved, and the method has higher application value in actual scenes.
S5, carrying out position adjustment on the primary candidate frame according to the offset vector to obtain standard position information of the primary candidate frame, optimizing the adjusted primary candidate frame by using a preset non-maximum suppression algorithm to obtain a standard candidate frame, and taking the standard candidate frame, the corresponding standard position information and the target defect as defect analysis results of the silicon wafer.
The size, proportion and position of the primary candidate frame generated by the preset area proposal network according to the noise reduction optimization diagram are fixed, however, the size and position of the defect area which is actually required to be detected are different, so that the primary candidate frame is required to be subjected to position adjustment according to the offset vector, thereby obtaining a more accurate frame and better covering the defect area.
In the embodiment of the present invention, the optimizing the adjusted primary candidate frame by using a preset non-maximum suppression algorithm to obtain a standard candidate frame includes:
sorting the primary candidate frames according to the defect probability values corresponding to the candidate features of the primary candidate frames to obtain probability sorting;
determining a primary candidate frame with the highest defect probability value in the probability ranking as a target candidate frame, and determining the probability ranking without the target candidate frame as a residual ranking;
calculating the cross ratio of the target candidate frame and the primary candidate frame in the residual sequencing one by one;
and removing the corresponding primary candidate frames with the cross ratio larger than the preset cross threshold in the residual sorting to obtain new probability sorting, and repeating all steps between determining the primary candidate frame with the highest defect probability value in the probability sorting as the target candidate frame and obtaining the new probability sorting until no primary candidate frame exists in the obtained new probability sorting, and determining all the target candidate frames as standard candidate frames.
In detail, the calculating the intersection ratio of the target candidate frame and the primary candidate frame in the residual sequence one by one includes:
calculating the intersection ratio of the target candidate frame and the primary candidate frame in the residual sequence by using the following intersection ratio formula:
wherein ,
representing the cross ratio, +.>
Representing a function of calculating the rectangular area, +.>
The target candidate box is represented by a frame,
representing the primary candidate box in the remaining ordering.
In detail, calculating the intersection ratio may help to evaluate the similarity between two candidate boxes, i.e. how much they overlap. By using the cross-over ratio, it is possible to determine which candidate frames have a higher degree of overlap and which candidate frames have a lower degree of overlap, thereby selectively excluding those frames that overlap with other candidate frames to avoid outputting a plurality of similar detection results. Therefore, the use of the overlap ratio in the non-maximum suppression algorithm to filter overlapping candidate frames can improve the accuracy and speed of target detection.
In the embodiment of the invention, the standard candidate frame, the corresponding standard position information and the target defect are used as the defect analysis result of the silicon wafer, wherein the target defect can determine the defect type of the silicon wafer, and the standard candidate frame and the corresponding standard position information can determine the specific position of the defect on the silicon wafer, so that the target requirement of the silicon wafer defect detection is met.
FIG. 4 is a functional block diagram of a system for detecting defects of a silicon wafer based on computer vision according to an embodiment of the present invention.
The silicon wafer defect detection system 100 based on computer vision according to the present invention can be installed in an electronic device. Depending on the functions implemented, the silicon wafer defect detection system 100 based on computer vision may include a picture acquisition module 101, a noise reduction optimization module 102, a candidate frame determination module 103, a candidate frame adjustment module 104, and a result analysis module 105. The module of the invention, which may also be referred to as a unit, refers to a series of computer program segments, which are stored in the memory of the electronic device, capable of being executed by the processor of the electronic device and of performing a fixed function.
In the present embodiment, the functions concerning the respective modules/units are as follows:
the picture acquisition module 101: carrying out camera calibration on preset microscopic equipment to obtain calibration equipment, and acquiring a silicon wafer picture of a preset silicon wafer by using the calibration equipment;
the noise reduction optimization module 102: dividing the silicon wafer picture by using a preset Markov random field division algorithm to obtain a defect division picture, carrying out bilateral noise reduction on the defect division picture by using a preset bilateral filtering function to obtain a bilateral filtering picture, and carrying out wiener noise reduction on the bilateral filtering picture by using a preset wiener filter to obtain a noise reduction optimization picture;
The determination candidate block module 103: generating a primary candidate frame according to the noise reduction optimization graph by using a preset area proposal network, extracting candidate features of the primary candidate frame, and mapping the candidate features to a preset probability space by using a preset classification sub-network to obtain defect categories and defect probability values corresponding to the candidate features;
the adjustment candidate box module 104: selecting a defect category with the defect probability value larger than a preset probability threshold as a target defect of the candidate feature, and mapping a primary candidate frame corresponding to the candidate feature to a preset vector space according to a preset regression sub-network to obtain an offset vector corresponding to the primary candidate frame;
the result analysis module 105: and carrying out position adjustment on the primary candidate frame according to the offset vector to obtain standard position information of the primary candidate frame, optimizing the adjusted primary candidate frame by using a preset non-maximum suppression algorithm to obtain a standard candidate frame, and taking the standard candidate frame, the corresponding standard position information and the target defect as defect analysis results of the silicon wafer.
In detail, each module in the silicon wafer defect detection system 100 based on computer vision in the embodiment of the present invention adopts the same technical means as the silicon wafer defect detection method based on computer vision described in fig. 1 to 3, and can produce the same technical effects, which are not repeated here.
It will be evident to those skilled in the art that the invention is not limited to the details of the foregoing illustrative embodiments, and that the present invention may be embodied in other specific forms without departing from the spirit or essential characteristics thereof.
The present embodiments are, therefore, to be considered in all respects as illustrative and not restrictive, the scope of the invention being indicated by the appended claims rather than by the foregoing description, and all changes which come within the meaning and range of equivalency of the claims are therefore intended to be embraced therein. Any reference signs in the claims shall not be construed as limiting the claim concerned.
The blockchain is a novel application mode of computer technologies such as distributed data storage, point-to-point transmission, consensus mechanism, encryption algorithm and the like. The Blockchain (Blockchain), which is essentially a decentralised database, is a string of data blocks that are generated by cryptographic means in association, each data block containing a batch of information of network transactions for verifying the validity of the information (anti-counterfeiting) and generating the next block. The blockchain may include a blockchain underlying platform, a platform product services layer, an application services layer, and the like.
The embodiment of the application can acquire and process the related data based on the artificial intelligence technology. Among these, artificial intelligence (Artificial Intelligence, AI) is the theory, method, technique and application system that uses a digital computer or a digital computer-controlled machine to simulate, extend and extend human intelligence, sense the environment, acquire knowledge and use knowledge to obtain optimal results.
Furthermore, it is evident that the word "comprising" does not exclude other elements or steps, and that the singular does not exclude a plurality. Multiple units or systems set forth in the system embodiments may also be implemented by one unit or system in software or hardware. The terms first, second, etc. are used to denote a name, but not any particular order.
Finally, it should be noted that the above-mentioned embodiments are merely for illustrating the technical solution of the present invention and not for limiting the same, and although the present invention has been described in detail with reference to the preferred embodiments, it should be understood by those skilled in the art that modifications and equivalents may be made to the technical solution of the present invention without departing from the spirit and scope of the technical solution of the present invention.
Claims (10)
1. The method for detecting the defects of the silicon wafer based on the computer vision is characterized by comprising the following steps of:
Carrying out camera calibration on preset microscopic equipment to obtain calibration equipment, and acquiring a silicon wafer picture of a preset silicon wafer by using the calibration equipment;
dividing the silicon wafer picture by using a preset Markov random field division algorithm to obtain a defect division picture, carrying out bilateral noise reduction on the defect division picture by using a preset bilateral filtering function to obtain a bilateral filtering picture, and carrying out wiener noise reduction on the bilateral filtering picture by using a preset wiener filter to obtain a noise reduction optimization picture;
generating a primary candidate frame according to the noise reduction optimization graph by using a preset area proposal network, extracting candidate features of the primary candidate frame, and mapping the candidate features to a preset probability space by using a preset classification sub-network to obtain defect categories and defect probability values corresponding to the candidate features;
selecting a defect category with the defect probability value larger than a preset probability threshold as a target defect of the candidate feature, and mapping a primary candidate frame corresponding to the candidate feature to a preset vector space according to a preset regression sub-network to obtain an offset vector corresponding to the primary candidate frame;
and carrying out position adjustment on the primary candidate frame according to the offset vector to obtain standard position information of the primary candidate frame, optimizing the adjusted primary candidate frame by using a preset non-maximum suppression algorithm to obtain a standard candidate frame, and taking the standard candidate frame, the corresponding standard position information and the target defect as defect analysis results of the silicon wafer.
2. The method for detecting a silicon wafer defect based on computer vision as defined in claim 1, wherein the step of calibrating the preset microscopic device with a camera to obtain calibration equipment comprises the steps of:
acquiring a preset black-and-white chessboard graph as a calibration plate of the microscopic equipment;
acquiring a calibration image of the calibration plate under a preset shooting condition by using a camera in the microscopic equipment;
detecting and positioning the calibration image by using a corner detection algorithm to obtain the position information of preset feature points;
according to preset calibration plate parameters, combining the position information, and calculating an internal parameter matrix of the camera by using a camera calibration algorithm;
according to the position information and the shooting conditions, calculating an external parameter matrix of the camera by using a preset camera pose estimation algorithm;
integrating the internal parameter matrix and the external parameter matrix into a camera parameter matrix, and determining microscopic equipment capable of correcting the shot picture by using the camera parameter matrix as calibration equipment.
3. The method for detecting a silicon wafer defect based on computer vision as defined in claim 1, wherein the dividing the silicon wafer picture by using a preset markov random field dividing algorithm to obtain a defect dividing map comprises:
Acquiring the region category of a pixel of a preset silicon wafer picture, and defining a potential function of the silicon wafer picture according to the region category;
calculating pixel region distribution of the silicon wafer picture according to the potential function;
the maximum pixel region distribution is determined as a defect segmentation map.
4. The method for detecting silicon wafer defects based on computer vision as recited in claim 3, wherein defining a potential function of the silicon wafer picture according to the region class comprises:
defining a potential function of the silicon wafer picture by using the following definition formula:
wherein ,
as a potential function +.>
Indicate->
Regional category to which each pixel belongs,>
indicate->
Regional category to which each pixel belongs,>
representing two unequal pixel sequence numbers, < ->
Indicate->
The individual pixels belong to->
Probability of->
Indicate->
And->
The individual pixels belong to->
and />
Is (are) joint probability->
Is the total number of pixel sequence numbers, +.>
Is a preset parameter for controlling smoothness.
5. The method for detecting a silicon wafer defect based on computer vision as defined in claim 3, wherein the obtaining the region category to which the pixel of the preset silicon wafer picture belongs comprises:
performing pixel coding on the silicon wafer picture to obtain a silicon wafer pixel;
Dividing the silicon wafer pixels into preset pixel areas;
and classifying the pixel areas by using a preset pixel label to obtain area categories.
6. The method for detecting a silicon wafer defect based on computer vision as set forth in claim 3, wherein said calculating a silicon wafer picture pixel area distribution according to said potential function comprises:
acquiring a pixel gray value of the silicon wafer picture;
calculating the distribution of pixel areas of the silicon wafer picture according to the pixel gray values by using the following probability distribution formula:
wherein ,
indicate->
Pixel gray value of individual pixels +.>
Representing the distribution of pixel areas of a silicon wafer picture, < >>
Represent the first
Regional category to which each pixel belongs,>
as a potential function +.>
Is a preset normalization constant.
7. The method for detecting a defect of a silicon wafer based on computer vision as set forth in claim 1, wherein said performing bilateral noise reduction on the defect segmentation map by using a preset bilateral filter function to obtain a bilateral filter map comprises:
obtaining a filtered pixel value corresponding to a pixel in the defect segmentation map by using the following bilateral filtering formula:
wherein ,
representing filtered pixel points subjected to bilateral filtering processing>
Corresponding filtered pixel values,/- >
Defective pixel point +.>
Corresponding defective pixel value,/->
Representing a preset filtering pixel point +.>
And defective pixel->
Spatial distance weight between->
Representing filtered pixel points +.>
And defective pixel->
Preset gray value similarity weight, +.>
Representing filtered pixel points +.>
Neighborhood of->
Is the weight sum of all gray value similarity weights;
and collecting the filtered pixel values into a double-sided filter map.
8. The method for detecting a silicon wafer defect based on computer vision according to any one of claims 1 to 7, wherein the step of performing wiener noise reduction on the bilateral filtering graph by using a preset wiener filter to obtain a noise reduction optimization graph comprises:
and carrying out wiener denoising on the bilateral filter graph by using a preset wiener filter by using the following wiener filter formula:
wherein ,
for the filtered noise reduction optimization map, +.>
For bilateral filter patterns, < >>
Is a preset band-pass filter +.>
Fourier transform of a predetermined degradation function, +.>
Signal power spectrum for bilateral filter patterns, +.>
Is the preset noise power.
9. The method for detecting a silicon wafer defect based on computer vision as set forth in claim 1, wherein the mapping the primary candidate frame corresponding to the candidate feature to a preset vector space according to a preset regression sub-network to obtain an offset vector corresponding to the primary candidate frame comprises:
Calculating an offset vector corresponding to the primary candidate frame by utilizing a boundary regression formula in the regression sub-network:
wherein ,
an abscissa representing the central coordinates of said primary candidate frame,/->
An ordinate representing the center coordinates of the primary candidate frame,/->
An abscissa representing the center coordinates of the reference frame corresponding to the primary candidate frame, +.>
Showing the ordinate of the center coordinates of the reference frame corresponding to the primary candidate frame, +.>
Representing the width of the primary candidate box, +.>
Representing the height, +.>
Representing the width of the reference frame corresponding to the primary candidate frame,/or->
Representing the height of the reference frame corresponding to the primary candidate frame,/>
For the abscissa position of the primary candidate frame relative to the corresponding reference frame,>
for the ordinate position of the primary candidate frame relative to the corresponding reference frame,>
for the width of the primary candidate frame relative to the corresponding reference frame,/a>
For the offset height of the primary candidate frame relative to the corresponding reference frame,/>
Is an offset vector.
10. A computer vision-based silicon wafer defect detection system, the system comprising:
and a picture acquisition module: carrying out camera calibration on preset microscopic equipment to obtain calibration equipment, and acquiring a silicon wafer picture of a preset silicon wafer by using the calibration equipment;
Noise reduction optimization module: dividing the silicon wafer picture by using a preset Markov random field division algorithm to obtain a defect division picture, carrying out bilateral noise reduction on the defect division picture by using a preset bilateral filtering function to obtain a bilateral filtering picture, and carrying out wiener noise reduction on the bilateral filtering picture by using a preset wiener filter to obtain a noise reduction optimization picture;
determining a candidate frame module: generating a primary candidate frame according to the noise reduction optimization graph by using a preset area proposal network, extracting candidate features of the primary candidate frame, and mapping the candidate features to a preset probability space by using a preset classification sub-network to obtain defect categories and defect probability values corresponding to the candidate features;
and (3) adjusting a candidate frame module: selecting a defect category with the defect probability value larger than a preset probability threshold as a target defect of the candidate feature, and mapping a primary candidate frame corresponding to the candidate feature to a preset vector space according to a preset regression sub-network to obtain an offset vector corresponding to the primary candidate frame;
and a result analysis module: and carrying out position adjustment on the primary candidate frame according to the offset vector to obtain standard position information of the primary candidate frame, optimizing the adjusted primary candidate frame by using a preset non-maximum suppression algorithm to obtain a standard candidate frame, and taking the standard candidate frame, the corresponding standard position information and the target defect as defect analysis results of the silicon wafer.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310521596.9A CN116228780B (en) | 2023-05-10 | 2023-05-10 | Silicon wafer defect detection method and system based on computer vision |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310521596.9A CN116228780B (en) | 2023-05-10 | 2023-05-10 | Silicon wafer defect detection method and system based on computer vision |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN116228780A true CN116228780A (en) | 2023-06-06 |
| CN116228780B CN116228780B (en) | 2023-07-07 |
Family
ID=86570098
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202310521596.9A Active CN116228780B (en) | 2023-05-10 | 2023-05-10 | Silicon wafer defect detection method and system based on computer vision |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116228780B (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117274241A (en) * | 2023-11-17 | 2023-12-22 | 四川赢信汇通实业有限公司 | Brake drum surface damage detection method and device based on rapid image analysis |
| CN117705816A (en) * | 2024-02-06 | 2024-03-15 | 西南交通大学 | Unmanned aerial vehicle-based steel rail surface defect detection method, system, equipment and medium |
| CN117705816B (en) * | 2024-02-06 | 2024-04-26 | 西南交通大学 | Unmanned aerial vehicle-based steel rail surface defect detection method, system, equipment and medium |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108010020A (en) * | 2017-11-29 | 2018-05-08 | 浙江华睿科技有限公司 | A kind of slideway detection method and device of silicon chip |
| CN110555842A (en) * | 2019-09-10 | 2019-12-10 | 太原科技大学 | Silicon wafer image defect detection method based on anchor point set optimization |
| CN111553929A (en) * | 2020-05-12 | 2020-08-18 | 重庆邮电大学 | Mobile phone screen defect segmentation method, device and equipment based on converged network |
| CN112184619A (en) * | 2020-08-25 | 2021-01-05 | 华北电力大学（保定） | Metal part surface defect detection method based on deep learning |
| CN114782310A (en) * | 2022-03-08 | 2022-07-22 | 江苏立导科技有限公司 | Surface defect detection method, device, equipment and storage medium |
-
2023
- 2023-05-10 CN CN202310521596.9A patent/CN116228780B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108010020A (en) * | 2017-11-29 | 2018-05-08 | 浙江华睿科技有限公司 | A kind of slideway detection method and device of silicon chip |
| CN110555842A (en) * | 2019-09-10 | 2019-12-10 | 太原科技大学 | Silicon wafer image defect detection method based on anchor point set optimization |
| CN111553929A (en) * | 2020-05-12 | 2020-08-18 | 重庆邮电大学 | Mobile phone screen defect segmentation method, device and equipment based on converged network |
| CN112184619A (en) * | 2020-08-25 | 2021-01-05 | 华北电力大学（保定） | Metal part surface defect detection method based on deep learning |
| CN114782310A (en) * | 2022-03-08 | 2022-07-22 | 江苏立导科技有限公司 | Surface defect detection method, device, equipment and storage medium |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117274241A (en) * | 2023-11-17 | 2023-12-22 | 四川赢信汇通实业有限公司 | Brake drum surface damage detection method and device based on rapid image analysis |
| CN117274241B (en) * | 2023-11-17 | 2024-02-09 | 四川赢信汇通实业有限公司 | Brake drum surface damage detection method and device based on rapid image analysis |
| CN117705816A (en) * | 2024-02-06 | 2024-03-15 | 西南交通大学 | Unmanned aerial vehicle-based steel rail surface defect detection method, system, equipment and medium |
| CN117705816B (en) * | 2024-02-06 | 2024-04-26 | 西南交通大学 | Unmanned aerial vehicle-based steel rail surface defect detection method, system, equipment and medium |
Also Published As
| Publication number | Publication date |
|---|---|
| CN116228780B (en) | 2023-07-07 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN115018828B (en) | Defect detection method for electronic component | |
| CN109978839B (en) | Method for detecting wafer low-texture defects | |
| CN108615226B (en) | Image defogging method based on generation type countermeasure network | |
| CN113570605B (en) | Defect detection method and system based on liquid crystal display panel | |
| CN114372983B (en) | Shielding box coating quality detection method and system based on image processing | |
| CN111429533B (en) | Camera lens distortion parameter estimation device and method | |
| CN111242026B (en) | Remote sensing image target detection method based on spatial hierarchy perception module and metric learning | |
| CN113608378B (en) | Full-automatic defect detection method and system based on LCD (liquid crystal display) process | |
| CN116228780B (en) | Silicon wafer defect detection method and system based on computer vision | |
| CN115841434A (en) | Infrared image enhancement method for gas concentration analysis | |
| CN112200848B (en) | Depth camera vision enhancement method and system under low-illumination weak-contrast complex environment | |
| CN115830004A (en) | Surface defect detection method, device, computer equipment and storage medium | |
| CN116664565A (en) | Hidden crack detection method and system for photovoltaic solar cell | |
| CN114674826A (en) | Visual detection method and detection system based on cloth | |
| CN115880683B (en) | Urban waterlogging ponding intelligent water level detection method based on deep learning | |
| CN109658405B (en) | Image data quality control method and system in crop live-action observation | |
| CN114742849B (en) | Leveling instrument distance measuring method based on image enhancement | |
| CN115546716A (en) | Binocular vision-based method for positioning fire source around power transmission line | |
| CN116309270A (en) | Binocular image-based transmission line typical defect identification method | |
| CN115205155A (en) | Distorted image correction method and device and terminal equipment | |
| CN113313678A (en) | Automatic sperm morphology analysis method based on multi-scale feature fusion | |
| CN117557565B (en) | Detection method and device for lithium battery pole piece | |
| CN116894841B (en) | Visual detection method for quality of alloy shell of gearbox | |
| CN115713501B (en) | Detection processing method and system suitable for blurred pictures of camera | |
| CN117115488B (en) | Water meter detection method based on image processing |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant | ||
| CP02 | Change in the address of a patent holder |
Address after: Room 102, Building 39, Software Park, Zhongguancun, Xinwu District, Wuxi City, Jiangsu Province, 214000 (Building D8-2) Patentee after: Jiangsu senbiao Technology Co.,Ltd. Address before: 2201, Wuxi Chamber of Commerce Building, No.1 Financial 8th Street, Economic Development Zone, Wuxi City, Jiangsu Province, 214000 Patentee before: Jiangsu senbiao Technology Co.,Ltd. |
|
| CP02 | Change in the address of a patent holder |