CN109102500A - A kind of automatic recognition system of chip package defect - Google Patents
A kind of automatic recognition system of chip package defect Download PDFInfo
- Publication number
- CN109102500A CN109102500A CN201810876863.3A CN201810876863A CN109102500A CN 109102500 A CN109102500 A CN 109102500A CN 201810876863 A CN201810876863 A CN 201810876863A CN 109102500 A CN109102500 A CN 109102500A
- Authority
- CN
- China
- Prior art keywords
- chip
- image
- defect
- module
- epoxy
- 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
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
- G01—MEASURING; TESTING
- G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
- G01N21/00—Investigating or analysing materials by the use of optical means, i.e. using sub-millimetre waves, infrared, visible or ultraviolet light
- G01N21/84—Systems specially adapted for particular applications
- G01N21/88—Investigating the presence of flaws or contamination
- G01N21/8851—Scan or image signal processing specially adapted therefor, e.g. for scan signal adjustment, for detecting different kinds of defects, for compensating for structures, markings, edges
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
-
- 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
- G01—MEASURING; TESTING
- G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
- G01N21/00—Investigating or analysing materials by the use of optical means, i.e. using sub-millimetre waves, infrared, visible or ultraviolet light
- G01N21/84—Systems specially adapted for particular applications
- G01N21/88—Investigating the presence of flaws or contamination
- G01N21/8851—Scan or image signal processing specially adapted therefor, e.g. for scan signal adjustment, for detecting different kinds of defects, for compensating for structures, markings, edges
- G01N2021/8854—Grading and classifying of flaws
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
- G01N21/00—Investigating or analysing materials by the use of optical means, i.e. using sub-millimetre waves, infrared, visible or ultraviolet light
- G01N21/84—Systems specially adapted for particular applications
- G01N21/88—Investigating the presence of flaws or contamination
- G01N21/8851—Scan or image signal processing specially adapted therefor, e.g. for scan signal adjustment, for detecting different kinds of defects, for compensating for structures, markings, edges
- G01N2021/8887—Scan or image signal processing specially adapted therefor, e.g. for scan signal adjustment, for detecting different kinds of defects, for compensating for structures, markings, edges based on image processing techniques
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Data Mining & Analysis (AREA)
- Life Sciences & Earth Sciences (AREA)
- General Engineering & Computer Science (AREA)
- Signal Processing (AREA)
- Evolutionary Biology (AREA)
- Bioinformatics & Computational Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Artificial Intelligence (AREA)
- Quality & Reliability (AREA)
- Evolutionary Computation (AREA)
- Health & Medical Sciences (AREA)
- Chemical & Material Sciences (AREA)
- Analytical Chemistry (AREA)
- Biochemistry (AREA)
- General Health & Medical Sciences (AREA)
- Immunology (AREA)
- Pathology (AREA)
- Image Processing (AREA)
Abstract
The present invention provides a kind of automatic recognition systems of chip package defect, include: chip image collecting module, and acquisition chip image obtains the original image that a frame includes multiple chips;Chip image preprocessing module obtains the chip image only containing a chip for pre-processing to collected original image;Chip image divides module, for being partitioned into epoxy regions from the chip image;Epoxy regions characteristic extracting module, for extracting the feature mix vector of reflection epoxy regions defect according to the epoxy regions;Epoxy resin defect recognition module is combined based on described eigenvector, is identified using defect classification of the classification learning algorithm to the epoxy resin.The automatic recognition system for the chip package defect that the present invention designs can in time, reliably be completed to improve the yield rate of chip package factory to the chip detection with epoxy resin defect, significantly reduce detection time and human cost.
Description
Technical field
The present invention relates to the technical field of chip package, especially a kind of automatic recognition system of chip package defect.
Background technique
Epoxy resin filling be one of important step of packaging technology, with the epoxide resin material of insulation be filled in wafer with
Between substrate, it is therefore an objective to reinforce chip structure, prevent from aoxidizing., can be unstable because of Glue dripping head in epoxy resin fill process, air pressure
The factors such as abnormal lead to problems such as epoxy resin askew or dispensing amount is inadequate in chip list millet cake, and these problems can make chip surface
Generate dispensing defect problem.
In semiconductor assembly and test workshop, asking for much reductions of the yield rate due to caused by epoxy resin defect is faced
Topic.Epoxy resin defect includes but is not limited to epoxy resin local perforations, epoxy resin part protrusion, epoxy resin excess edge
The defects of spilling or epoxy resin ullage emargintion, causes yield rate reduction due to the defect, and yield loss rate is high
Up to 0.04%.
The defects detection of existing epoxy resin be by manually after epoxy resin complete underfill process after its
After his packaging and testing process, before packaging factory, artificial detection is carried out to module is completed.But this artificial detection lacks timely
Property, in addition, reliability and validity are poor since degree of dependence of this method to operator is very high, and need very high
Cost of labor.
Summary of the invention
In view of the above-mentioned problems, the present invention is intended to provide a kind of automatic recognition system of chip package defect, can eliminate people
Work relies on, and improves the reliability and validity that epoxy resin detects and cost of labor is greatly reduced.
The purpose of the present invention is realized using following technical scheme:
A kind of automatic recognition system of chip package defect is provided, includes:
Chip image collecting module, acquisition chip image obtain the original image that a frame includes multiple chips;
Illumination compensation module, for when the chip image collecting module is acquired chip image, to be collected
Chip carries out illumination compensation;
Chip image preprocessing module, for pre-processing to collected original image, the pretreatment includes cutting
Divide processing and enhanced processing, obtains the chip image only containing a chip;The chip image includes wafer, is centered around crystalline substance
The epoxy regions on first periphery and the substrate outside epoxy regions;
Chip image divides module, for being partitioned into epoxy regions from the chip image;
Epoxy regions characteristic extracting module, for extracting reflection epoxy resin according to the epoxy regions
The feature mix vector of area defects;
Epoxy resin defect recognition module, based on obtained described eigenvector combination, using the classification of support vector machines
Learning algorithm identifies the defect classification of the epoxy resin.
Preferably, the automatic recognition system further includes image storage module, includes online storage element and training storage
Memory cell, it is described to be used to store the obtained chip image after pre-processing in line storage element, and by corresponding chip number
Also it is stored, the chip number includes the combination of one of chip type, code name and batch, two or three;It is described
Sample storage unit is used to store the multiclass sample image of Training Support Vector Machines.
Preferably, the automatic recognition system further includes recognition result output module, is sent to for defect recognition result
Back-end processor, so that defect when back-end processor carries out epoxy resin filling according to chip adjusts production process
It is whole.
Preferably, the chip image collecting module includes high-definition camera, and the video camera is using face battle array arrangement
Monochrome CCD sensor camera.
Preferably, the illumination compensation module includes multiple flat light source modules, and the flat light source module is by red
LED light takes stationary arrangement to combine to obtain with blue LED lamp;It further include distance adjusting unit, for adjusting the video camera mirror
Head is at a distance from the chip.
Preferably, the detailed process that epoxy regions are partitioned into from the chip image are as follows: will obtain first
The chip image carry out gray proces, gray level image corresponding with chip image and gradient image are obtained, by its normalizing
Change handles to obtain by normalized gray level image and gradient image, and then establishes grey gradient probability confederate matrix;Then it calculates
It finds partition threshold (z, s) to be split the grey gradient probability confederate matrix, obtains gray value of image greater than z, gradient value
Submatrix greater than s;Wherein, z is the partition value of chip image gray level, and s is the segmentation level of chip image gradient grade;Finally,
The corresponding pixel of element is split the chip image in the submatrix obtained according to segmentation, obtains the epoxy resin
Region.
Preferably, the calculation formula of the partition threshold are as follows:
In formula, (z, s) is optimum segmentation threshold values, the pijFor grey gradient probability confederate matrix PCIn the i-th row jth column
Element value, pass throughIt can be calculated;xijIt represents so that gray value is equal to i and gradient in image
The number of pixel of the value equal to j;HdFor the maximum gray level of setting;TdFor the greatest gradient grade of setting;Z is correspondence image
The partition value of gray level, s are the segmentation level of correspondence image gradient grade.
The invention has the benefit that a kind of automatic recognition system of chip package defect provided by the invention, by right
Collected chip image is handled to obtain epoxy regions, and then extracts the feature mix vector of epoxy regions simultaneously
It is identified according to defect classification of the vector to epoxy regions, and then optimization is adjusted to production process, in time,
It is reliable to complete to detect the chip package with epoxy resin defect, in this way, improving the yield rate of chip package factory, pole
The earth reduces detection time cost and human cost.
Detailed description of the invention
The present invention will be further described with reference to the accompanying drawings, but the embodiment in attached drawing is not constituted to any limit of the invention
System, for those of ordinary skill in the art, without creative efforts, can also obtain according to the following drawings
Other attached drawings.
Fig. 1 is the process flow that the chip package defect inspection system is added in a preferred embodiment of the invention
Figure;
Fig. 2 is the comprising modules of the automatic recognition system of chip package defect described in a preferred embodiment of the invention
Figure.
Specific embodiment
In conjunction with following application scenarios, the invention will be further described.
Referring to Fig. 1 and Fig. 2, a kind of automatic recognition system of chip package defect is provided in the present embodiment, includes:
Chip image collecting module, acquisition chip image obtain the original image that a frame includes multiple chips;
Illumination compensation module, for when the chip image collecting module is acquired chip image, to be collected
Chip carries out illumination compensation;
Chip image preprocessing module, for pre-processing to collected original image, the pretreatment includes cutting
Divide processing and enhanced processing, obtains the chip image only containing a chip;The chip image includes wafer, is centered around crystalline substance
The epoxy regions on first periphery and the substrate outside epoxy regions;
Chip image divides module, for being partitioned into epoxy regions from the chip image;
Epoxy regions characteristic extracting module, for extracting reflection epoxy resin according to the epoxy regions
The feature mix vector of area defects;
Epoxy resin defect recognition module, based on obtained described eigenvector combination, using the classification of support vector machines
Learning algorithm identifies the defect classification of the epoxy resin.
In the present embodiment, the automatic recognition system further includes image storage module, includes online storage element and instruction
Practice storage element, it is described to be used to store the obtained chip image after pre-processing in line storage element, and by corresponding chip
Number is also stored, and the chip number includes the combination of one of chip type, code name and batch, two or three;
The sample storage unit is used to store the multiclass sample image of Training Support Vector Machines.
The multiclass sample image includes multiple chip figure image sets, specifically: the chip figure of epoxy resin local perforations
Image set, epoxy resin chip figure image set and asphalt mixtures modified by epoxy resin that locally chip figure image set outstanding, epoxy resin excess edge overflow
The chip figure image set of rouge ullage emargintion;The epoxy resin local perforations refer to epoxy regions memory in chip image
In the hole for being not filled by or filling very thin epoxy resin;Locally protrusion refers to epoxy resin region in chip image to the epoxy resin
There is the epoxy resin protrusion that filling is excessive or filling is very thick but does not overflow in domain;Epoxy resin excess edge spilling refers to
It excessively causes to overflow in the presence of filling in epoxy regions in chip image;The epoxy resin ullage emargintion refers to chip
Lack of fill causes epoxy regions edge notch occur in epoxy regions in image.
It concentrates epoxy regions defect classification different according to multiple chip images, the multiclass sample image is passed through certainly
It is split in dynamic identifying system and obtains corresponding feature mix vector with feature extraction, by the feature mix vector and figure
It is sent into support vector machines as corresponding defect classification and carries out off-line training, obtain the classifier of training completion;In view of this reality
Apply the defect of epoxy resin filling when only considering above-mentioned common 4 kinds of encapsulation in example, and just training one between every two kinds of defects
Classifier, therefore 6 classifiers of project training to 4 kinds of defects carry out classification identification, are distinguished in classifier by ballot
The On-line testing to feature mix vector which kind of defect belonged to.
In the present embodiment, the automatic recognition system further includes recognition result output module, is sent out for defect recognition result
Send to back-end processor so that back-end processor according to chip carry out epoxy resin filling when defect to production process into
Row adjustment.
In the present embodiment, the chip image collecting module includes high-definition camera, and the video camera is using face battle array row
The monochrome CCD sensor camera of cloth.
In the present embodiment, the illumination compensation module includes multiple flat light source modules, the flat light source module by
Red LED lamp takes stationary arrangement to combine to obtain with blue LED lamp;It further include distance adjusting unit, for adjusting the camera shooting
Machine camera lens is at a distance from the chip.
In the present embodiment, the distance adjusting unit is according to the camera lens parameter of selection come described in adaptive calculating
Camera lens are realized and the distance of the two are adjusted at a distance from the chip;Wherein, the camera lens with
The calculation formula of the distance of the chip are as follows:
In formula, CdFor the vertical range of camera lens and determinand surface;F is the focal length of camera lens;Q is camera shooting
The maximum ring of machine camera lens;D allows disperse circular diameter for the CCD camera inner sensor;G is the prospect of camera lens
It is deep;K (p) is the illumination compensation factor.
In this preferred embodiment, it is provided with illumination compensation module, illumination compensation is carried out when to chip image collecting, and design
Distance adjusting unit, by reasonably being adjusted at a distance from the chip to the camera lens, so that Image Acquisition
Hardware view is optimised, and collected original image quality is good, precision is high.
It is described that cutting processing and enhanced processing are carried out to collected original image in the present embodiment, obtain the core
The detailed process of picture are as follows: firstly, the filter window using 3*3 carries out median filtering to the original image, will filter
The image arrived carries out binary conversion treatment, obtains bianry image;Cutting is carried out to bianry image based on anatomic element parser, is examined
Can not only partial noise have been eliminated but also can retain overall profile information by considering opening operation, therefore use rectangular configuration element to two-value processing
Later bianry image carries out opening operation processing, and rectangular element and background are carried out cutting to reach, obtain multiple initial chips
Image;Then, closed operation processing then using rectangular configuration element is carried out to multiple initial chip images, respectively to reach connection simultaneously
The purpose of smooth chip local edge;The image after closed operation is processed is finally amplified processing using algorithm to obtain
Chip image.
In the present embodiment, the detailed process that epoxy regions are partitioned into from the chip image are as follows: first will
The obtained chip image carries out gray proces, gray level image corresponding with chip image and gradient image is obtained, by it
Normalized is obtained by normalized gray level image and gradient image, and then establishes grey gradient probability confederate matrix;Then
It calculates searching partition threshold (z, s) to be split the grey gradient probability confederate matrix, obtains gray value of image greater than z, ladder
Angle value is greater than the submatrix of s;Wherein, z is the partition value of chip image gray level, and s is the segmentation level of chip image gradient grade;Most
Afterwards, the corresponding pixel of element is split the chip image in the submatrix obtained according to segmentation, obtains discrete ring
Oxygen resin area, then dilation operation is successively carried out to discrete epoxy regions based on the algorithm of mathematical morphology, so that ring
The connection of oxygen resin area, obtains the epoxy regions in image.
In the present embodiment, the calculation formula of the partition threshold are as follows:
In formula, (z, s) is optimum segmentation threshold values, the pijFor grey gradient probability confederate matrix PCIn the i-th row jth column
Element value, pass throughIt can be calculated;xijIt represents so that gray value is equal to i and gradient in image
The number of pixel of the value equal to j;HdFor the maximum gray level of setting;TdFor the greatest gradient grade of setting;Z is correspondence image
The partition value of gray level, s are the segmentation level of correspondence image gradient grade.
In this preferred embodiment, in order to be partitioned into the epoxy regions from chip image, calculated using threshold segmentation
Method handles image, has comprehensively considered the gray scale and gradient value of the original image, to design the calculating of partition threshold
Formula carries out the segmentation of chip image according to the partition threshold being calculated, so that segmentation precision is high.
In the present embodiment, the feature mix vector of the reflection epoxy regions defect includes epoxy regions
Geometrical characteristic, gray feature and textural characteristics;And the specific mode of seeking of this three category feature are as follows:
(1) it the Extraction of Geometrical Features of epoxy regions: extracts for expressing the several of epoxy regions geometry
What feature, specific formula for calculation are as follows:
In formula, MtFor the geometrical characteristic functional value that represent epoxy regions geometry;N is epoxy resin area
The total number of pixel in domain;lBFor the long side length of the epoxy regions minimum circumscribed rectangle;lSFor the asphalt mixtures modified by epoxy resin
The bond length of rouge region minimum circumscribed rectangle;L is the number of pixel on the epoxy regions boundary;Y is asphalt mixtures modified by epoxy resin
The set that pixel is constituted in rouge region;(u, v) is any pixel in the epoxy regions;U is respectively represented
The abscissa value of pixel (u, v) in the epoxy regions;V is the vertical of pixel (u, v) in the epoxy regions
Coordinate value;(uo,v0) be the epoxy regions barycentric coodinates;F (u, v) is pixel (u, v) in epoxy regions
Pixel value.
(2) extraction of gray feature: the gray level image with the epoxy regions is first obtained, corresponding grey scale is then obtained
The gray level of image and corresponding histogram, are calculated the probability that each gray value occurs, and then obtain the gray level image
Overall intensity mean value and gray level entropy, and gray average and gray level entropy are merged to obtain gray scale spy according to certain amalgamation mode
Sign;
(3) extraction of textural characteristics: firstly, generating gray level co-occurrence matrixes corresponding with the epoxy regions;It is based on
Gray level co-occurrence matrixes calculate its textural characteristics:
In formula, wnThe functional value of textural characteristics is corresponded to for the epoxy regions;K is epoxy regions corresponding grey scale
The gray level of figure;R is the gray value of any pixel point (u, v) in epoxy regions;T be epoxy regions in and pixel
(u, v) distance is the gray value of another pixel in the neighborhood of d;According to the number that pixel (r, t) occurs, construct about distance
The gray level co-occurrence matrixes W of d;Distance d is selected according to the periodic distribution of texture;W (r, t) is r-1 in gray level co-occurrence matrixes W
Row t-1 arranges corresponding element value.
In this preferred embodiment, propose subsequent scarce to epoxy resin to realize to epoxy regions progress feature extraction
Sunken identification has comprehensively considered more representational a few category feature values of reflection epoxy regions defect, respectively epoxy
Geometrical characteristic, gray feature and the Gradient Features of resin area;And based on repeatedly test, geometrical characteristic and texture are devised
The calculation formula of feature, so that the geometrical characteristic of epoxy regions and textural characteristics have obtained effective description, and whole
On obtained combination of eigenvectors can accurately and fast reflect the defect classifications of the epoxy regions.
In the present embodiment, based on obtained described eigenvector combination, the defect classification of the epoxy resin is known
Other detailed process are as follows: 6 classifiers completed according to above-mentioned training carry out Classification and Identification come the feature vector obtained to extraction,
To obtain corresponding epoxy resin defect;Described eigenvector combination is respectively fed to 6 classification that above-mentioned training is completed
In device carry out defect classification, it is contemplated that the classifier is only two-value classifier, wherein described eigenvector any two not
Ballot classification is carried out between same defect type, so carries out the model that the classification of global multiclass defect is converted into multiple two classification
It solves, number of votes obtained statistics is carried out to the classification results of 6 classifiers;The poll that last real-time statistics obtain, when counting
The defect classification most to poll is that described eigenvector combines the identified epoxy resin defect type come out.
In the present embodiment, the automatic recognition system is in the attachment process of chip, the attachment process to include: core
Piece attachment, underfill, other encapsulation and testing process, artificial detection and packaging factory;The automatic recognition system is used
In the chip image collecting to underfill epoxy, and then analysis handles to obtain the defect classification of filling epoxy resin, if
It was found that defect, then chip will be corresponded to by, which stopping, being sent into subsequent handling processing.
In this preferred embodiment, the automatic recognition system of a kind of chip package defect provided, by collected core
Picture is handled to obtain epoxy regions, and then extracts the feature mix vector of epoxy regions and according to the vector
The defect classification of epoxy regions is identified, and then optimization is adjusted to production process, in time, reliably to complete
Chip package detection with epoxy resin defect is significantly reduced in this way, improving the yield rate of chip package factory
Detection time cost and human cost.
Finally it should be noted that the above embodiments are merely illustrative of the technical solutions of the present invention, rather than the present invention is protected
The limitation of range is protected, although explaining in detail referring to preferred embodiment to the present invention, those skilled in the art are answered
Work as analysis, it can be with modification or equivalent replacement of the technical solution of the present invention are made, without departing from the reality of technical solution of the present invention
Matter and range.
Claims (7)
1. a kind of automatic recognition system of chip package defect, which is characterized in that include:
Chip image collecting module, acquisition chip image obtain the original image that a frame includes multiple chips;
Illumination compensation module, for treating acquisition chip when the chip image collecting module is acquired chip image
Carry out illumination compensation;
Chip image preprocessing module, for pre-processing to collected original image, the pretreatment includes at cutting
Reason and enhanced processing, obtain the chip image only containing a chip;The chip image includes wafer, is centered around wafer week
The epoxy regions on side and the substrate outside epoxy regions;
Chip image divides module, for being partitioned into epoxy regions from the chip image;
Epoxy regions characteristic extracting module, for extracting reflection epoxy regions according to the epoxy regions
The feature mix vector of defect;
Epoxy resin defect recognition module, based on obtained described eigenvector combination, using the classification learning of support vector machines
Algorithm identifies the defect classification of the epoxy resin.
2. a kind of automatic recognition system of chip package defect according to claim 1, which is characterized in that the automatic knowledge
Other system further includes image storage module, includes online storage element and training storage element, and the online storage element is used
In the chip image that storage obtains after pretreatment, and corresponding chip number is also stored, the chip number packet
Include the combination of one of chip type, code name and batch, two or three;The sample storage unit is for storing training branch
Hold the multiclass sample image of vector machine.
3. a kind of automatic recognition system of chip package defect according to claim 1, which is characterized in that the automatic knowledge
Other system further includes recognition result output module, back-end processor is sent to for defect recognition result, so that at rear end
Defect when reason device carries out epoxy resin filling according to chip is adjusted production process.
4. a kind of automatic recognition system of chip package defect according to claim 1, which is characterized in that the chip figure
As the monochrome CCD sensor camera that acquisition module includes high-definition camera, and the video camera is arranged using face battle array.
5. a kind of automatic recognition system of chip package defect according to claim 1, which is characterized in that the illumination is mended
Repaying module includes multiple flat light source modules, and the flat light source module takes fixed row by red LED lamp and blue LED lamp
Column combination obtains;It further include distance adjusting unit, for adjusting the camera lens at a distance from the chip.
6. a kind of automatic recognition system of chip package defect according to claim 1, which is characterized in that described from described
The detailed process of epoxy regions is partitioned into chip image are as follows: first carry out the obtained chip image at gray scale
Reason, obtains gray level image corresponding with chip image and gradient image, its normalized is obtained by normalized ash
Image and gradient image are spent, and then establishes grey gradient probability confederate matrix;Then it calculates and finds partition threshold (z, s) for the ash
Gradient probability confederate matrix is split, and obtains gray value of image greater than z, gradient value is greater than the submatrix of s;Wherein, z is chip
The partition value of image gray levels, s are the segmentation level of chip image gradient grade;Finally, element in the submatrix obtained according to segmentation
Corresponding pixel is split the chip image, obtains the epoxy regions.
7. a kind of automatic recognition system of chip package defect according to claim 6, which is characterized in that the segmentation valve
The calculation formula of value are as follows:
In formula, (z, s) is optimum segmentation threshold values, the pijFor grey gradient probability confederate matrix PCIn the i-th row jth column member
Element value, passes throughIt can be calculated;xijIt represents so that gray value is equal to i and gradient value etc. in image
In the number of the pixel of j;HdFor the maximum gray level of setting;TdFor the greatest gradient grade of setting;Z is correspondence image gray scale
The partition value of grade, s are the segmentation level of correspondence image gradient grade.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810876863.3A CN109102500A (en) | 2018-08-03 | 2018-08-03 | A kind of automatic recognition system of chip package defect |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810876863.3A CN109102500A (en) | 2018-08-03 | 2018-08-03 | A kind of automatic recognition system of chip package defect |
Publications (1)
Publication Number | Publication Date |
---|---|
CN109102500A true CN109102500A (en) | 2018-12-28 |
Family
ID=64848367
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201810876863.3A Pending CN109102500A (en) | 2018-08-03 | 2018-08-03 | A kind of automatic recognition system of chip package defect |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN109102500A (en) |
Cited By (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111127571A (en) * | 2019-12-03 | 2020-05-08 | 歌尔股份有限公司 | Small sample defect classification method and device |
CN113484325A (en) * | 2021-07-05 | 2021-10-08 | 广东奥普特科技股份有限公司 | Wafer dispensing defect detection device and wafer dispensing defect detection method |
CN114943738A (en) * | 2022-07-25 | 2022-08-26 | 武汉飞恩微电子有限公司 | Sensor packaging curing adhesive defect identification method based on visual identification |
Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102621150A (en) * | 2012-03-23 | 2012-08-01 | 南京航空航天大学 | Airplane skin damage identification method based on gray level co-occurrence matrix and support vector machine |
CN108198766A (en) * | 2017-12-29 | 2018-06-22 | 英特尔产品(成都)有限公司 | Chip package defect inspection and processing method, system and storage device |
-
2018
- 2018-08-03 CN CN201810876863.3A patent/CN109102500A/en active Pending
Patent Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102621150A (en) * | 2012-03-23 | 2012-08-01 | 南京航空航天大学 | Airplane skin damage identification method based on gray level co-occurrence matrix and support vector machine |
CN108198766A (en) * | 2017-12-29 | 2018-06-22 | 英特尔产品(成都)有限公司 | Chip package defect inspection and processing method, system and storage device |
Non-Patent Citations (3)
Title |
---|
周德龙等: "基于灰度-梯度共生矩阵模型的最大熵阈值处理算法", 《小型微型计算机系统》 * |
巢渊: "基于机器视觉的半导体芯片表面缺陷在线检测关键技术研究", 《万方数据》 * |
陈后金等: "基于灰度-梯度共生矩阵的钢轨表面缺陷检测方法", 《北京交通大学学报》 * |
Cited By (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111127571A (en) * | 2019-12-03 | 2020-05-08 | 歌尔股份有限公司 | Small sample defect classification method and device |
US11836967B2 (en) | 2019-12-03 | 2023-12-05 | Goertek, Inc. | Method and device for small sample defect classification and computing equipment |
CN111127571B (en) * | 2019-12-03 | 2023-12-29 | 歌尔股份有限公司 | Small sample defect classification method and device |
CN113484325A (en) * | 2021-07-05 | 2021-10-08 | 广东奥普特科技股份有限公司 | Wafer dispensing defect detection device and wafer dispensing defect detection method |
CN114943738A (en) * | 2022-07-25 | 2022-08-26 | 武汉飞恩微电子有限公司 | Sensor packaging curing adhesive defect identification method based on visual identification |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN108765371B (en) | Segmentation method of unconventional cells in pathological section | |
CN111161243B (en) | Industrial product surface defect detection method based on sample enhancement | |
CN109378052B (en) | The preprocess method and system of image labeling | |
CN109829914A (en) | The method and apparatus of testing product defect | |
US7215829B2 (en) | Method and system for object recognition using fractal map | |
CN109767422A (en) | Pipe detection recognition methods, storage medium and robot based on deep learning | |
CN109102500A (en) | A kind of automatic recognition system of chip package defect | |
CN110085709B (en) | Full-automatic counting and counting system for LED images | |
CN108416814B (en) | Method and system for quickly positioning and identifying pineapple head | |
CN112085024A (en) | Tank surface character recognition method | |
CN109685045A (en) | A kind of Moving Targets Based on Video Streams tracking and system | |
CN116310845B (en) | Intelligent monitoring system for sewage treatment | |
US7062079B2 (en) | Method and system for image segmentation | |
CN108830175A (en) | Iris image local enhancement methods, device, equipment and storage medium | |
CN110827312A (en) | Learning method based on cooperative visual attention neural network | |
CN108664839A (en) | A kind of image processing method and equipment | |
CN112990103A (en) | String mining secondary positioning method based on machine vision | |
CN111783819A (en) | Improved target detection method based on region-of-interest training on small-scale data set | |
CN116228740A (en) | Small sample chip appearance defect detection method and detection system based on improved YOLOv5 | |
CN114781514A (en) | Floater target detection method and system integrating attention mechanism | |
CN116228678A (en) | Automatic identification and processing method for chip packaging defects | |
CN112348018A (en) | Digital display type instrument reading identification method based on inspection robot | |
CN114040116B (en) | Plastic mould good product monitoring feedback system | |
CN116843581B (en) | Image enhancement method, system, device and storage medium for multi-scene graph | |
Hu et al. | Automatic detection of pecan fruits based on Faster RCNN with FPN in orchard |
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 | ||
RJ01 | Rejection of invention patent application after publication |
Application publication date: 20181228 |
|
RJ01 | Rejection of invention patent application after publication |