CN108447050A - A kind of Surface Flaw dividing method based on super-pixel - Google Patents
A kind of Surface Flaw dividing method based on super-pixel Download PDFInfo
- Publication number
- CN108447050A CN108447050A CN201810186200.9A CN201810186200A CN108447050A CN 108447050 A CN108447050 A CN 108447050A CN 201810186200 A CN201810186200 A CN 201810186200A CN 108447050 A CN108447050 A CN 108447050A
- Authority
- CN
- China
- Prior art keywords
- image
- pixel
- super
- defect
- block
- 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
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/23—Clustering 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
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/40—Analysis of texture
- G06T7/41—Analysis of texture based on statistical description of texture
- G06T7/45—Analysis of texture based on statistical description of texture using co-occurrence matrix computation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/60—Analysis of geometric attributes
- G06T7/62—Analysis of geometric attributes of area, perimeter, diameter or volume
-
- 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
- G06T2207/30164—Workpiece; Machine component
-
- 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
- G06T2207/30168—Image quality inspection
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- General Engineering & Computer Science (AREA)
- Computing Systems (AREA)
- Evolutionary Computation (AREA)
- Bioinformatics & Computational Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- Evolutionary Biology (AREA)
- Mathematical Physics (AREA)
- Probability & Statistics with Applications (AREA)
- Quality & Reliability (AREA)
- Geometry (AREA)
- Image Analysis (AREA)
- Investigating Materials By The Use Of Optical Means Adapted For Particular Applications (AREA)
Abstract
A kind of Surface Flaw dividing method based on super-pixel, its key points of the technical solution are that:Finished work surface image is subjected to image block, obtains the subimage block of m*m pixel size;The energy value and entropy of subimage block gray level co-occurrence matrixes are calculated, one is chosen and is used as standard master drawing, remaining subimage block energy value and entropy in contrast, are filtered out containing defective image;Using super-pixel SLIC algorithms by the image segmentation containing defect be super-pixel;Super-pixel is clustered using spectral clustering NJW algorithms, obtains defect Segmentation image;Calculate the center of gravity and region area of defect, and according to image subsection block position and camera and actual size relationship, be calculated defect workpiece surface practical specific location.The method of the present invention from super-pixel level calculate and be reduced for high Resolution and Large Size image huge calculation amount pixel-based, computational efficiency higher, while also playing inhibiting effect for the interference of background texture, noise in image.
Description
Technical field
The present invention relates to a kind of Surface Flaw dividing method based on super-pixel, belong to workpiece surface quality detection and
Image processing field.
Background technology
Quality or subsequent job step important of the workpiece surface quality to product.Especially in high precision part
When processing, due to chip, machine vibration, tool wear and clamping carrying etc., workpiece surface often will produce a variety of lack
It falls into, product orientation accuracy decline, service life is caused to reduce.Because Surface Flaw is small, processing site interference is strong and
Off-line type manually inspects the proposition of the present situation and digital factory concept for the detection unstable quality brought, the people of off-line type by random samples
Work detection has not been suitable for reality.
Therefore the surface quality detection method based on machine vision becomes the hot spot of research.Surface Flaw Detection is general
It is divided into the detection algorithm based on edge detection algorithm and based on similar area.The representative method of first kind edge detection is that differential is calculated
Sub- edge detection algorithm and surface fitting edge detection algorithm.Method is simple but noise inhibiting ability is poor.Second class region detection
Representative method Threshold segmentation and cluster.But due to the randomness and complexity of image, to the more demanding of image preprocessing.
Above method is calculated both for image single pixel, and the size for usually handling image is 256*256, number
It is very big according to measuring.With the raising of industrial digitalization level, high real-time, high reliability high rate burst communication algorithm become machine
A unusual critical issue in the application of device vision.Super-pixel segmentation segments the image into multiple images block, it is intended to utilize the mankind
Vision be as unit of image block, rather than go to understand as unit of pixel image this according to removing processing image.And by image
After being divided into super-pixel, it is used for subsequent processing using the feature of super-pixel as parameter, greatly reduces the redundancy of pixel, carries
The high treatment effeciency of image.
Invention content
To solve the above-mentioned problems, the present invention proposes a kind of Surface Flaw extracting method based on super-pixel.Work
Part surface defect image is screened, and will be contained defective image and is split using super-pixel, is ensured consistent in super-pixel
Property, and defect area is extracted by cluster, and positioned.
It is special in order to realize that the target, the present invention propose a kind of Surface Flaw dividing method based on super-pixel
Sign is, includes the following steps:
Step S1 workpiece surface image blocks, finished work surface image is directly acquired by industrial camera, to obtaining
Image carry out image block, so that image is divided into the subimage block of m*m pixel size;
Step S2 chooses one and does not have defective subimage block as standard pattern, and calculates the energy of its gray level co-occurrence matrixes
Magnitude and entropy will be larger with standard master drawing energy value and entropy difference by comparing the energy value and entropy of other images
Image is defined as containing defective image;
Step S3 super-pixel generates, and will be judged in step S2 containing defective image point using super-pixel SLIC algorithms
It is segmented into multiple irregular super-pixel;
Step S4 spectral clusterings are classified, and are classified to super-pixel block using spectral clustering NJW algorithms, are obtained the super picture of background texture
Plain block is one kind, and defect super-pixel block is the image of one kind, completes defect Segmentation, and carry out binaryzation, defect part 0, the back of the body
Scape texture is 1.
The positioning of step S5 defects calculates center of gravity and the area of defect according to the binary image after completion defect Segmentation
Domain area, and according to the focal length of image subsection block position and camera when step S1 image blocks, defect is calculated in workpiece
The practical specific location on surface.
The step S1 workpiece surface image blocks include the following steps:Use industrial camera acquisition workpiece surface to be detected
Complete image, and selective positioning point is labeled;According to the installation site and focal length of industrial camera, workpiece surface image ruler is obtained
The very little relationship between resolution ratio;It is the subimage block of m*m that workpiece surface complete image to be detected, which is divided into resolution ratio, and right
Every image is numbered as Nij, line number of the wherein i expression image subsections in original image, j expression subgraphs are in original image
In columns.
The step S2 calculates the energy value of gray level co-occurrence matrixes and entropy includes the following steps:Step S21 chooses one
Do not have a defective subimage block, calculate 0 °, 45 °, 90 °, 135 ° of four directions, the gray level co-occurrence matrixes that distance is 1 pixel, carry
The energy value and entropy of gray level co-occurrence matrixes are taken out, and calculates the average value of energy value and entropy, is asked using statistical concepts
Go out energy value, the entropy value range of defect-free surface image;Remaining subimage block is pressed step S21 the methods by step S22
The average value not subimage block in standard master drawing value range has been defined as scarce by the average value for calculating energy value and entropy
Sunken image.
The step S3 super-pixel generation includes the following steps:The figure containing blemish surface that step S31 will classify in step S2
The original sub image block of picture is transformed into the spaces HSI from rgb space, and extracts I component gray-scale map;Step S32 initialization K is poly-
Class central point Ck=[Ik,xk,yk], cluster centre space of points distanceEach pixel in the region of 2S*2S is calculated to arrive
The distance of cluster centre pointWhereinPixel is assigned to most
On the central point of short d, cluster process is completed;At the end of step S33 clusters, it will produce and be not belonging to the identical connection point of its cluster centre
The lonely pixel of amount founds point, is corrected using connected component method, these isolated pixels is made to be assigned on nearest cluster centre point.
The step S4 spectral clusterings include the following steps:The all pixels in each super-pixel that step S3 is generated are averaged
Brightness value of the brightness value as the super-pixel carries out spectral clustering as sample characteristics using NJW algorithms, clusters number by
GAP statistic laws determine.
The step S5 defect locations include the following steps:Defect area occupied area size in bianry image is calculated first
With centroid pixel position,Wherein S is defect area pixel summation, and R is defect area;Then according to son
Position coordinates of the image block in original image calculate the location of pixels in original workpiece surface image of defect,Defect is calculated in workpiece finally by the relationship between workpiece surface picture size and resolution ratio
Physical location and area.
The beneficial effects of the invention are as follows:The present invention using gray level co-occurrence matrixes to complete workpiece surface piecemeal subimage block into
Row screening judges that the subimage block containing defective target extracts the subsequent accurate extraction of progress by preliminary, reduces for complete
The huge calculation amount that whole workpiece surface image high-resolution large scale generates.Then super-pixel segmentation algorithm is used to be lacked according to difference
Sunken shape feature is split, and clusters each super-pixel block using spectral clustering to obtain defect bianry image.
Finally by the calculating to defect area center of gravity and area, the physical location of defect area is obtained.Intersect at traditional Pixel-level
Other extraction algorithm, super-pixel segmentation algorithm computational efficiency higher simplify complicated calculating process, while being directed in image and carrying on the back
The interference of scape texture, noise also plays inhibiting effect.
Description of the drawings
Fig. 1 is the flow chart of the Surface Flaw extracting method based on super-pixel;
Fig. 2 is finished work surface image piecemeal schematic diagram;
Fig. 3 is Surface Flaw image superpixel segmentation schematic diagram;
Fig. 4 is bianry image schematic diagrames after Surface Flaw segmentation;
Fig. 5 is the flow chart of energy value and entropy extraction of the step S2 based on gray level co-occurrence matrixes;
Fig. 6 is the flow chart that step S3 super-pixel generates.
Specific implementation mode
To make the objectives, technical solutions, and advantages of the present invention clearer, below in conjunction with specific embodiment, and reference
Attached drawing, the present invention will be described in further detail.
Fig. 1 is the Surface Flaw extracting method flow chart the present invention is based on super-pixel.As shown in Figure 1, described be based on
The Surface Flaw extracting method of super-pixel includes the following steps:
Step S1:Workpiece surface image block directly acquires finished work surface image by industrial camera, to obtaining
Image carry out image block, so that image is divided into the subimage block of m*m pixel size;
Workpiece surface complete image to be detected is acquired using industrial camera, the size of gained image is 2048*1536.Selection
The a certain length of side of workpiece is labeled as anchor point, and obtains workpiece surface figure according to industrial camera, installation site and focal length
As the relationship between size and resolution ratio, then show that a pixel represents on workpiece actual size as 0.0218mm*
0.0218mm.It is the subimage block of 256*256 that workpiece surface complete image to be detected, which is divided into resolution ratio, obtains 6 rows 8 row
Image subsection block, and N is numbered to every image method as shown in Figure 2ij, row of the wherein i expression image subsections in original image
Number, j indicate columns of the subgraph in original image;
Step S2:Choosing one does not have defective subimage block as standard pattern, and calculates its gray level co-occurrence matrixes
Energy value and entropy.It is larger with standard master drawing energy value and entropy difference by comparing the energy value and entropy of other images
Image is defined as containing defective image;
The step S2 further comprises the steps:
Step S21 chooses one and does not have defective subimage block as standard sample, calculates 0 °, 45 °, 90 °, 135 ° four
Direction, distance be 1 pixel standard sample gray level co-occurrence matrixes, obtain 4 gray level co-occurrence matrixes extract its energy value and
Entropy simultaneously calculates the energy value of the standard sample and the average value of entropy.It is total that standard sample gray scale is found out using statistical concepts
The energy value of raw matrix, entropy value range;
Step S22 is by remaining subimage block by the average value for calculating energy value and entropy described in step S21.Not by average value
Subimage block in standard master drawing value range, is defined as defective image.
Step S3 super-pixel generates, and contains defect image for what is judged in step 2, to each imagery exploitation super-pixel
SLIC algorithms will be divided into multiple irregular super-pixel;
The step S3 further comprises the steps:
The original sub image block for the image containing blemish surface classified in step 2 is transformed into HSI by step S31 from rgb space
Space, and extract I component gray-scale map;
Step S32 initializes K cluster centre point Ck=[Ik,xk,yk], cluster centre space of points distance
Wherein K is input parameter, the number of super-pixel to be generated;IkFor each cluster centre pixel brightness value in image;xk,ykFor in this
Itself pixel coordinate of heart point.In order to generate the approximately equal super-pixel of size, equably minute by cluster centre points initial K
Cloth calculates initially according to the total number N of K and whole image pixel away from the cluster centre space of points on the mesh point of image space
DistanceUsing the pixel corresponding to the image minimal gradient value of 3*3 neighborhoods as new initial cluster center point.
Calculate distance of each pixel to cluster centre point in the region of 2S*2SWhereinPixel is assigned on the central point of most short d, completes cluster process.
At the end of step S33 clusters, the vertical point of lonely pixel for being not belonging to the identical connected component of its cluster centre is will produce, is used
Connected component method is corrected, these isolated pixels is made to be assigned on nearest cluster centre point.
Step S4 spectral clusterings are classified, and are classified to super-pixel block using spectral clustering NJW algorithms, are obtained the super picture of background texture
Plain block is one kind, and defect super-pixel block is the defect Segmentation image of one kind.All pictures in each super-pixel that step S3 is generated
Brightness value of the average brightness value of element as the super-pixel carries out spectral clustering using NJW algorithms, gathers as sample characteristics
Class number is determined by GAP statistic laws;
The positioning of step S5 defects calculates the center of gravity of defect, and according to step according to the bianry image after defect Segmentation
The focal length of image subsection block position and camera when 1 image block, be calculated defect workpiece surface practical specific location;
Defect area occupied area size and centroid pixel position in bianry image are calculated first,Its
Middle Q is defect area pixel summation, and R is defect area;
Then the position coordinates according to subimage block in original image, calculate defect in original workpiece surface image
In location of pixels.
Actual bit of the defect in workpiece is calculated finally by the relationship between workpiece surface picture size and resolution ratio
It sets and area;
Fig. 2 is the schematic diagram that complete workpiece surface image block is image subsection, is numbered according to image subsection, subgraph is calculated
As the position coordinates of block on the original image, be then calculated subimage block workpiece surface physical location;Fig. 3 is workpiece
Surface defect image super-pixel segmentation schematic diagram, super-pixel SLIC algorithms are drawn according to the brightness value and spatial coordinate location of pixel
Multiple irregular super-pixel block are separated, boundary line is divided and is bonded defect area boundary well, therefore super-pixel can be according to defect
Edges of regions is divided, and good segmentation effect has been obtained;Fig. 4 is bianry image schematic diagrames after Surface Flaw segmentation,
Defect area is extracted, while removing the feed texture on finished surface during spectral clustering, it is suppressed that difference processing line
The interference extracted to defect is managed, illustrates the validity of the method for the present invention.
Claims (6)
1. a kind of Surface Flaw dividing method based on super-pixel, which is characterized in that include the following steps:
Step S1 workpiece surface image blocks, directly acquire finished work surface image, to the figure of acquisition by industrial camera
As carrying out image block, image is made to be divided into the subimage block of m*m pixel size;
Step S2 chooses one and does not have defective subimage block as standard pattern, and calculates the energy value of its gray level co-occurrence matrixes
And entropy, by comparing the energy value and entropy of other images, by the image larger with standard master drawing energy value and entropy difference,
It is defined as containing defective image;
Step S3 super-pixel generates, and is containing defective image segmentation by what is judged in step S2 using super-pixel SLIC algorithms
Multiple irregular super-pixel;
Step S4 spectral clusterings are classified, and are classified to super-pixel block using spectral clustering NJW algorithms, are obtained background texture super-pixel block
For one kind, defect super-pixel block is the image of one kind, completes defect Segmentation, and carry out binaryzation, defect part 0, background line
Reason is 1.
The positioning of step S5 defects calculates the center of gravity and area surface of defect according to the binary image after completion defect Segmentation
Product, and according to the focal length of image subsection block position and camera when step S1 image blocks, defect is calculated in workpiece surface
Practical specific location.
2. the Surface Flaw dividing method according to claim 1 based on super-pixel, which is characterized in that the step
S1 workpiece surface image blocks include the following steps:Workpiece surface complete image to be detected is acquired using industrial camera, and is selected
Anchor point is labeled;According to the installation site and focal length of industrial camera, obtain between workpiece surface picture size and resolution ratio
Relationship;It is the subimage block of m*m that workpiece surface complete image to be detected, which is divided into resolution ratio, and is compiled to every image
Number be Nij, line number of the wherein i expression image subsections in original image, columns of the j expression subgraphs in original image.
3. the Surface Flaw dividing method according to claim 1 based on super-pixel, which is characterized in that the step
S2 calculates the energy value of gray level co-occurrence matrixes and entropy includes the following steps:Step S21 chooses one and does not have defective subgraph
Block, calculate 0 °, 45 °, 90 °, 135 ° of four directions, distance be 1 pixel gray level co-occurrence matrixes, extract gray level co-occurrence matrixes
Energy value and entropy, and calculate the average value of energy value and entropy, defect-free surface image found out using statistical concepts
Energy value, entropy value range;Remaining subimage block is calculated energy value and entropy by step S22 by step S21 the methods
Average value the average value not subimage block in standard master drawing value range is defined as defective image.
4. the Surface Flaw dividing method according to claim 1 based on super-pixel, which is characterized in that the step
The generation of S3 super-pixel includes the following steps:Step S31 is by the original sub image block for the image containing blemish surface classified in step S2
The spaces HSI are transformed into from rgb space, and extract I component gray-scale map;Step S32 initializes K cluster centre point Ck=[Ik,
xk,yk], cluster centre space of points distanceCalculate in the region of 2S*2S each pixel to cluster centre point away from
FromWhereinPixel is assigned on the central point of most short d,
Complete cluster process;At the end of step S33 cluster, will produce be not belonging to the identical connected component of its cluster centre lonely pixel it is vertical
Point is corrected using connected component method, these isolated pixels is made to be assigned on nearest cluster centre point.
5. the Surface Flaw dividing method according to claim 1 based on super-pixel, which is characterized in that the step
S4 spectral clusterings include the following steps:The average brightness value for all pixels in each super-pixel that step S3 is generated surpasses picture as this
The brightness value of element carries out spectral clustering, clusters number is determined by GAP statistic laws as sample characteristics using NJW algorithms.
6. the Surface Flaw dividing method according to claim 1 based on super-pixel, which is characterized in that the step
S5 defect locations include the following steps:Defect area occupied area size and centroid pixel position in bianry image are calculated first,Wherein S is defect area pixel summation, and R is defect area;Then according to subimage block in original image
In position coordinates, calculate the location of pixels in original workpiece surface image of defect,Finally
Physical location and area of the defect in workpiece are calculated by the relationship between workpiece surface picture size and resolution ratio.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810186200.9A CN108447050A (en) | 2018-03-07 | 2018-03-07 | A kind of Surface Flaw dividing method based on super-pixel |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810186200.9A CN108447050A (en) | 2018-03-07 | 2018-03-07 | A kind of Surface Flaw dividing method based on super-pixel |
Publications (1)
Publication Number | Publication Date |
---|---|
CN108447050A true CN108447050A (en) | 2018-08-24 |
Family
ID=63193476
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201810186200.9A Pending CN108447050A (en) | 2018-03-07 | 2018-03-07 | A kind of Surface Flaw dividing method based on super-pixel |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN108447050A (en) |
Cited By (12)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109277327A (en) * | 2018-09-26 | 2019-01-29 | 芜湖常瑞汽车部件有限公司 | A kind of detection method of part defective products |
CN109493339A (en) * | 2018-11-20 | 2019-03-19 | 北京嘉恒中自图像技术有限公司 | A kind of casting inner surface gas hole defect detection method based on endoscopic imaging |
CN109658381A (en) * | 2018-11-16 | 2019-04-19 | 华南理工大学 | A kind of copper face defect inspection method of the flexible IC package substrate based on super-pixel |
CN109872303A (en) * | 2019-01-16 | 2019-06-11 | 北京交通大学 | Surface defect visible detection method, device and electronic equipment |
CN110111308A (en) * | 2019-04-12 | 2019-08-09 | 国网江苏省电力有限公司电力科学研究院 | Carbon fiber composite core wire ray image processing method, defect inspection method, device, equipment and computer storage medium |
CN110189297A (en) * | 2019-04-18 | 2019-08-30 | 杭州电子科技大学 | A kind of magnetic material open defect detection method based on gray level co-occurrence matrixes |
CN111429395A (en) * | 2019-01-08 | 2020-07-17 | 鸿富锦精密电子(成都)有限公司 | Tool life prediction method, device and computer storage medium |
CN113176200A (en) * | 2021-04-19 | 2021-07-27 | 西安交通大学 | Friction evolution measuring instrument and measuring method for rough surface morphology |
CN113341881A (en) * | 2021-06-17 | 2021-09-03 | 无锡互盛智能科技有限公司 | Control system for numerical control machine tool |
CN115049835A (en) * | 2022-08-16 | 2022-09-13 | 众烁精密模架(南通)有限公司 | Data preprocessing method based on die-casting die defect identification |
CN115063413A (en) * | 2022-08-04 | 2022-09-16 | 宁波鑫芯微电子科技有限公司 | Feature extraction method for abnormal data of super-large-scale wafer |
CN117237298A (en) * | 2023-09-15 | 2023-12-15 | 广州乾丰印花有限公司 | Printed fabric defect inspection method, device and computing equipment |
Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102346851A (en) * | 2011-11-04 | 2012-02-08 | 西安电子科技大学 | Image segmentation method based on NJW (Ng-Jordan-Weiss) spectral clustering mark |
CN103413316A (en) * | 2013-08-24 | 2013-11-27 | 西安电子科技大学 | SAR image segmentation method based on superpixels and optimizing strategy |
CN106093066A (en) * | 2016-06-24 | 2016-11-09 | 安徽工业大学 | A kind of magnetic tile surface defect detection method based on the machine vision attention mechanism improved |
CN106530303A (en) * | 2016-11-01 | 2017-03-22 | 陕西科技大学 | Spectral clustering-based color image fast segmentation method |
CN109584247A (en) * | 2018-11-20 | 2019-04-05 | 陕西师范大学 | It is a kind of based on semi-supervised super-pixel spectral clustering color image segmentation method |
CN111583279A (en) * | 2020-05-12 | 2020-08-25 | 重庆理工大学 | Super-pixel image segmentation method based on PCBA |
-
2018
- 2018-03-07 CN CN201810186200.9A patent/CN108447050A/en active Pending
Patent Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102346851A (en) * | 2011-11-04 | 2012-02-08 | 西安电子科技大学 | Image segmentation method based on NJW (Ng-Jordan-Weiss) spectral clustering mark |
CN103413316A (en) * | 2013-08-24 | 2013-11-27 | 西安电子科技大学 | SAR image segmentation method based on superpixels and optimizing strategy |
CN106093066A (en) * | 2016-06-24 | 2016-11-09 | 安徽工业大学 | A kind of magnetic tile surface defect detection method based on the machine vision attention mechanism improved |
CN106530303A (en) * | 2016-11-01 | 2017-03-22 | 陕西科技大学 | Spectral clustering-based color image fast segmentation method |
CN109584247A (en) * | 2018-11-20 | 2019-04-05 | 陕西师范大学 | It is a kind of based on semi-supervised super-pixel spectral clustering color image segmentation method |
CN111583279A (en) * | 2020-05-12 | 2020-08-25 | 重庆理工大学 | Super-pixel image segmentation method based on PCBA |
Non-Patent Citations (7)
Title |
---|
X1188822: "SLIC超像素分割算法和目前超像素算法的比较", 《HTTPS://WWW.DOC88.COM/P-7458620899350.HTML?R=1》 * |
南柄飞等: "基于SLIC0融合纹理信息的超像素分割方法", 《仪器仪表学报》 * |
周友行等: "工件表面缺陷图像检测中的自适应聚类", 《表面技术》 * |
崔建斌等: "确定最佳聚类数的二阶差分统计法", 《安徽大学学报(自然科学版)》 * |
邹旭华等: "基于改进的相似度度量的谱聚类图像分割方法", 《计算机工程与应用》 * |
马逐曦: "基于超像素的平面铣削工件表面缺陷视觉检测研究", 《中国优秀硕士学位论文全文数据库 (工程科技Ⅰ辑)》 * |
高娜: "基于超像素谱聚类的图像分割方法", 《中国优秀硕士学位论文全文数据库 (工程科技Ⅰ辑)》 * |
Cited By (17)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109277327A (en) * | 2018-09-26 | 2019-01-29 | 芜湖常瑞汽车部件有限公司 | A kind of detection method of part defective products |
CN109658381A (en) * | 2018-11-16 | 2019-04-19 | 华南理工大学 | A kind of copper face defect inspection method of the flexible IC package substrate based on super-pixel |
CN109493339B (en) * | 2018-11-20 | 2022-02-18 | 北京嘉恒中自图像技术有限公司 | Endoscope imaging-based method for detecting defects of pores on inner surface of casting |
CN109493339A (en) * | 2018-11-20 | 2019-03-19 | 北京嘉恒中自图像技术有限公司 | A kind of casting inner surface gas hole defect detection method based on endoscopic imaging |
CN111429395A (en) * | 2019-01-08 | 2020-07-17 | 鸿富锦精密电子(成都)有限公司 | Tool life prediction method, device and computer storage medium |
CN109872303A (en) * | 2019-01-16 | 2019-06-11 | 北京交通大学 | Surface defect visible detection method, device and electronic equipment |
CN110111308A (en) * | 2019-04-12 | 2019-08-09 | 国网江苏省电力有限公司电力科学研究院 | Carbon fiber composite core wire ray image processing method, defect inspection method, device, equipment and computer storage medium |
CN110111308B (en) * | 2019-04-12 | 2022-08-23 | 国网江苏省电力有限公司电力科学研究院 | Method and device for processing radiation image of carbon fiber composite core wire |
CN110189297A (en) * | 2019-04-18 | 2019-08-30 | 杭州电子科技大学 | A kind of magnetic material open defect detection method based on gray level co-occurrence matrixes |
CN110189297B (en) * | 2019-04-18 | 2021-02-19 | 杭州电子科技大学 | Magnetic material appearance defect detection method based on gray level co-occurrence matrix |
CN113176200A (en) * | 2021-04-19 | 2021-07-27 | 西安交通大学 | Friction evolution measuring instrument and measuring method for rough surface morphology |
CN113341881A (en) * | 2021-06-17 | 2021-09-03 | 无锡互盛智能科技有限公司 | Control system for numerical control machine tool |
CN115063413A (en) * | 2022-08-04 | 2022-09-16 | 宁波鑫芯微电子科技有限公司 | Feature extraction method for abnormal data of super-large-scale wafer |
CN115049835A (en) * | 2022-08-16 | 2022-09-13 | 众烁精密模架(南通)有限公司 | Data preprocessing method based on die-casting die defect identification |
CN115049835B (en) * | 2022-08-16 | 2022-11-29 | 众烁精密模架(南通)有限公司 | Data preprocessing method based on die-casting die defect identification |
CN117237298A (en) * | 2023-09-15 | 2023-12-15 | 广州乾丰印花有限公司 | Printed fabric defect inspection method, device and computing equipment |
CN117237298B (en) * | 2023-09-15 | 2024-05-14 | 广州乾丰印花有限公司 | Printed fabric defect inspection method, device and computing equipment |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN108447050A (en) | A kind of Surface Flaw dividing method based on super-pixel | |
CN109829891B (en) | Magnetic shoe surface defect detection method based on dense generation of antagonistic neural network | |
CN111383209B (en) | Unsupervised flaw detection method based on full convolution self-encoder network | |
CN109978839B (en) | Method for detecting wafer low-texture defects | |
CN104990925B (en) | One kind is based on gradient multi thresholds optimization defect inspection method | |
CN110286126A (en) | A kind of wafer surface defects subregion area detecting method of view-based access control model image | |
CN106446894B (en) | A method of based on outline identification ball-type target object location | |
CN107248159A (en) | A kind of metal works defect inspection method based on binocular vision | |
CN109540925B (en) | Complex ceramic tile surface defect detection method based on difference method and local variance measurement operator | |
CN108985337A (en) | A kind of product surface scratch detection method based on picture depth study | |
CN106780526A (en) | A kind of ferrite wafer alligatoring recognition methods | |
CN112085675B (en) | Depth image denoising method, foreground segmentation method and human motion monitoring method | |
CN107490583A (en) | A kind of intermediate plate defect inspection method based on machine vision | |
CN110648330B (en) | Defect detection method for camera glass | |
CN114820625B (en) | Automobile top block defect detection method | |
CN115100206B (en) | Printing defect identification method for textile with periodic pattern | |
CN110807763A (en) | Method and system for detecting ceramic tile surface bulge | |
CN115100199A (en) | Method for detecting wafer low-texture defects | |
CN115063620B (en) | Bit layering based Roots blower bearing wear detection method | |
CN114463314A (en) | Wafer defect detection method and system based on color difference shadow model | |
CN114972216A (en) | Construction method and application of texture surface defect detection model | |
CN115953407B (en) | Semiconductor equipment maintenance system based on computer vision | |
US20110164129A1 (en) | Method and a system for creating a reference image using unknown quality patterns | |
CN113506246A (en) | Concrete 3D printing component fine detection method based on machine vision | |
CN117635609B (en) | Visual inspection method for production quality of plastic products |
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 | ||
WD01 | Invention patent application deemed withdrawn after publication | ||
WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20180824 |