CN111161213A - Industrial product defect image classification method based on knowledge graph - Google Patents
Industrial product defect image classification method based on knowledge graph Download PDFInfo
- Publication number
- CN111161213A CN111161213A CN201911250467.0A CN201911250467A CN111161213A CN 111161213 A CN111161213 A CN 111161213A CN 201911250467 A CN201911250467 A CN 201911250467A CN 111161213 A CN111161213 A CN 111161213A
- Authority
- CN
- China
- Prior art keywords
- defect
- network
- image
- graph
- batch
- 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 130
- 238000000034 method Methods 0.000 title claims abstract description 29
- 238000000605 extraction Methods 0.000 claims description 19
- 235000019580 granularity Nutrition 0.000 claims description 9
- 238000011176 pooling Methods 0.000 claims description 9
- 238000013145 classification model Methods 0.000 claims description 7
- 230000002708 enhancing effect Effects 0.000 claims description 7
- 230000006870 function Effects 0.000 claims description 6
- 230000008569 process Effects 0.000 claims description 6
- 239000002356 single layer Substances 0.000 claims description 6
- 230000009466 transformation Effects 0.000 claims description 6
- 238000006243 chemical reaction Methods 0.000 claims description 4
- 239000010410 layer Substances 0.000 claims description 4
- 239000011159 matrix material Substances 0.000 claims description 4
- 238000004422 calculation algorithm Methods 0.000 claims description 3
- 230000002950 deficient Effects 0.000 claims description 3
- 238000005516 engineering process Methods 0.000 claims description 3
- 230000003993 interaction Effects 0.000 claims description 3
- 238000005457 optimization Methods 0.000 claims description 3
- 230000008859 change Effects 0.000 claims description 2
- 238000013135 deep learning Methods 0.000 abstract description 6
- 238000009776 industrial production Methods 0.000 abstract description 4
- 238000013527 convolutional neural network Methods 0.000 abstract 1
- 238000013528 artificial neural network Methods 0.000 description 6
- 238000004519 manufacturing process Methods 0.000 description 6
- 238000010586 diagram Methods 0.000 description 5
- 230000000694 effects Effects 0.000 description 2
- 101150064138 MAP1 gene Proteins 0.000 description 1
- 238000013473 artificial intelligence Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000007635 classification algorithm Methods 0.000 description 1
- 101150077939 mapA gene Proteins 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
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
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- General Physics & Mathematics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- General Engineering & Computer Science (AREA)
- Life Sciences & Earth Sciences (AREA)
- Evolutionary Computation (AREA)
- Artificial Intelligence (AREA)
- Health & Medical Sciences (AREA)
- Computational Linguistics (AREA)
- Evolutionary Biology (AREA)
- Quality & Reliability (AREA)
- Bioinformatics & Computational Biology (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Bioinformatics & Cheminformatics (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses a method for classifying industrial product defect images based on a knowledge graph, which can classify industrial product defects by combining experience knowledge in industrial production and the image characteristics of the defects, and overcomes the defect that the defects are classified only by the characteristics of the images of a conventional convolutional neural network. The method can greatly improve the accuracy of deep learning in the classification of industrial product defects while reducing the dependence on defect samples.
Description
Technical Field
The invention belongs to the field of computer vision, and particularly relates to a knowledge graph-based industrial product defect image classification method.
Background
In industrial production, in order to optimize the production process and improve the production quality, the method has very important significance for correctly classifying the defects in the production process of the product. Generally, the classification of defects of industrial products must be effectively classified by workers trained professionally, which is inefficient and often difficult to ensure. On one hand, enterprises pay high labor cost for the image classification, and on the other hand, the image classification process is very tedious and the post rotation rate is high. With the introduction of industry 4.0 and the development of artificial intelligence techniques, image classification techniques typified by deep learning have been widely used. Compared with the traditional classification algorithm, the method has the advantages that the complex characteristic engineering is not needed in deep learning, the adaptability is strong, the conversion is easy, the production flexibility can be met, the technical requirement of production iteration can be quickly met, and the method has strong advantages in the application of industrial product defect image classification.
Unlike image classification in natural scenes, however, the defect classification rules for industrial images are often based on factory production experience, not just on the characteristics of the images themselves. In addition, in many cases, the difference between different types of defects is small, and the types of defects are often unbalanced, which brings great challenges for deep learning and defect classification applied to industrial images.
Disclosure of Invention
Aiming at the defects of the prior art, the invention provides a method for classifying industrial product defect images based on a knowledge graph, which combines the prior knowledge of industrial production defects and the characteristics of the defect images to train end to end by adopting a mode of combining a graph convolution neural network and a convolution neural network, and can greatly improve the accuracy of deep learning in classification of industrial product defects while reducing the dependence on defect samples.
The purpose of the invention is realized by the following technical scheme:
s1: respectively creating a defect map library X, a label library Y corresponding to each defect map and an additional attribute vector V corresponding to the defect type;
s2: and enhancing the defect map by utilizing an image enhancement technology to obtain an enhanced defect map library X'.
S3: constructing a defect image feature extraction network, and then continuously acquiring image batches X from the defect map library X' enhanced by S2BAnd its corresponding tag yB. Grouping x defective picturesBInputting the defect image feature extraction network to obtain image features of the image batch under different granularities, and normalizing to obtain normalized image features eB；
The defect image feature extraction network comprises a backbone network and a layered bilinear pooling network which are trained by Imagenet and used for extracting the VGG network features. After features u, v and w of the image under different granularities are extracted by a backbone network for extracting the VGG network features, cross-layer interaction of local characteristics is captured by utilizing a bilinear pooling network. The bilinear pooling network H has the following form:
H(u，v，w)＝PTconcat(UTu*VTv，UTu*WTw，VTv*WTw)
wherein P is a classification matrix, U, V and W are projection matrices of U, V and W respectively, and H (U, V and W) is the image characteristics of the image batches under different granularities.
S4: a plan embedding manner;
e obtained from S3BThe feature of each defect in (1) and the additional attribute vector V corresponding to the defect type in S1 are embedded into an undirected graph, thereby obtaining a graph batch g of batch image featuresB；
S5: constructing a graph convolution network for graph feature extraction, and dividing the graph batch g in S4BInputting the graph into the graph volume network to obtain a graph batch gBCharacteristic o ofB；
S6: for o in S5BFeature transformation is carried out to obtain transformed feature o'B。
S7: constructing a single-layer linear discriminant network, and then carrying out feature e in S3BO 'obtained from S6'BThe output of the discrimination network is a defect image batch xBCorresponding discrimination tag y'B；
S8: calculating defect image batch x from loss functionBTrue tag y ofBAnd discrimination tag y 'obtained in S7'BLoss between ld；
S9: simultaneous optimization of l with back-propagation algorithmdThe defect image feature extraction network, the graph convolution network and the discrimination network are carried out until the step ldConverging to obtain the whole defect mapA classification model of the image;
s10: and performing the feature processing of the processes from S3 to S6 on the defect picture to be classified and the additional attribute vector corresponding to the defect type, and inputting the processed additional attribute vector into the classification model of the defect image obtained in S9, wherein the defect image feature extraction network, the graph convolution network and the single-layer linear discrimination network in S3-S6 are optimized networks, and finally the defect type to which the defect belongs is obtained.
Further, the enhancing the defect samples by using the image enhancement technique in S2 includes randomly turning over the negative samples, shifting, changing contrast, saturation, brightness, and adding noise; meanwhile, when data is enhanced, the meaning of the defects is ensured not to change.
Further, the feature transformation of S6 is calculated by the following rule:
R(x)＝Conv1D(f(x))
wherein R is an inverse conversion function of the graph convolution characteristic, Conv1D is a one-dimensional convolution network, and f is a linear rectification function with a specific form as follows:
further, the loss l in S8dCalculated from the cross entropy loss, for which the discriminant class probability m and the defect class label n are entered, n ∈ Bj，
Wherein α is the weight of the class probability m, j is the number of defect classes, BjA vector set of defect class labels.
The invention has the following beneficial effects:
compared with the existing deep learning classification method, the method for classifying the defect images of the industrial products based on the knowledge-graph provided by the invention utilizes the graph convolution neural network and the convolution neural network to simultaneously integrate the priori knowledge of industrial production and the characteristics of the images to classify the defects of the images, and solves the defect that the conventional convolution neural network only classifies the defects from the characteristics of the images. The method improves the classification robustness, can greatly reduce the dependence on the defect sample in the actual classification task, can be widely applied to the task of industrial product defect images, and has good universality and universality.
Drawings
FIG. 1 is a method and flow diagram according to the present invention;
FIG. 2 is a defect classification diagram flow and defect feature display diagram without knowledge-graph basis according to the present invention;
FIG. 3 is a defect classification diagram flow and defect feature display diagram based on knowledge graph according to the present invention;
Detailed Description
The present invention will be described in detail below with reference to the accompanying drawings and preferred embodiments, and the objects and effects of the invention will become more apparent. It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.
As shown in fig. 1, the method for classifying defect images of industrial products based on knowledge-graph of the present invention comprises the following steps:
s1: respectively creating a defect map library X, a label library Y corresponding to each defect map and an additional attribute vector V corresponding to the defect type;
s2: and enhancing the defect map by utilizing an image enhancement technology to obtain an enhanced defect map library X'.
S3: constructing a defect image feature extraction network, and then continuously acquiring image batches X from the defect map library X' enhanced by S2BAnd its corresponding tag yB. Grouping x defective picturesBInputting the defect image feature extraction network to obtain image features of the image batch under different granularities, and normalizing to obtain normalized image features eB；
The defect image feature extraction network comprises a backbone network and a layered bilinear pooling network which are trained by Imagenet and used for extracting the VGG network features. After features u, v and w of the image under different granularities are extracted by a backbone network for extracting the VGG network features, cross-layer interaction of local characteristics is captured by utilizing a bilinear pooling network. The bilinear pooling network H has the following form:
H(u，v，w)＝PTconcat(UTu*VTv，UTu*WTw，VTv*WTw)
wherein P is a classification matrix, U, V and W are projection matrices of U, V and W respectively, and H (U, V and W) is the image characteristics of the image batches under different granularities.
S4: a plan embedding manner;
e obtained from S3BThe feature of each defect in (1) and the additional attribute vector V corresponding to the defect type in S1 are embedded into an undirected graph, thereby obtaining a graph batch g of batch image featuresB；
For graph batch gBIn each graph G, a vertex is established according to the dimension of the feature and the type of the defect, an edge connected with the dimension point is added to the vertex of each defect type, and for the graph G, the following features are provided:
the number of vertices of G is S3, the dimension of each feature + the number of defect types
The number of edges of G is S3, and dimension x number of defect types of each feature is obtained
G defect point vertex characteristic is additional attribute vector V in S1
G feature point vertex feature being GBi，gBiIs gBCorresponding to a single defect image feature in (1).
S5: constructing a graph convolution network for graph feature extraction, and dividing the graph batch g in S4BInputting the graph into the graph volume network to obtain a graph batch gBCharacteristic o ofB(ii) a For the graph convolution neural network, the following form is given:
wherein,
is the normalized binary adjacency matrix of the input graph, X' is the characteristic of the input graph of the graph, W is the weight of the graph convolution network layer, and Z is the output of the graph convolution network.
S6: for o in S5BFeature transformation is carried out to obtain transformed feature o'B. The feature transformation is computed by means of the following rules:
R(x)＝Conv1D(f(x))
wherein R is an inverse conversion function of the graph convolution characteristic, Conv1D is a one-dimensional convolution network, and f is a linear rectification function with a specific form as follows:
s7: constructing a single-layer linear discriminant network, and then carrying out feature e in S3BO 'obtained from S6'BThe output of the discrimination network is a defect image batch xBCorresponding discrimination tag y'B；
S8: calculating defect image batch x from loss functionBTrue tag y ofBAnd discrimination tag y 'obtained in S7'BLoss between ld(ii) a Loss ldCalculated from the cross entropy loss, for which the discriminant class probability m and the defect class label n are entered, n ∈ Bj，
Wherein α is the weight of the class probability m, j is the number of defect classes, BjA vector set of defect class labels.
S9: simultaneous optimization of l with back-propagation algorithmdThe defect image feature extraction network, the graph convolution network and the discrimination network are carried out until the step ldConverging to finally obtain a classification model of the whole defect image;
s10: and performing the feature processing of the processes from S3 to S6 on the defect picture to be classified and the additional attribute vector corresponding to the defect type, and inputting the processed additional attribute vector into the classification model of the defect image obtained in S9, wherein the defect image feature extraction network, the graph convolution network and the single-layer linear discrimination network in S3-S6 are optimized networks, and finally the defect type to which the defect belongs is obtained.
As shown in fig. 3, a flowchart for defect image classification is shown. As shown in fig. 2 and 3, under the condition of not being based on the knowledge graph and being based on the knowledge graph, the same batch of defect data is classified, and the defect features in the classification process are analyzed and found, the defect features obtained in the process of directly classifying the defects by using the neural network model can hardly be separated on the feature space, and when the knowledge graph corresponding to the defects is added into the classification model, the distinguishing of the features of the defects on the feature space becomes easy, which shows that the defect image classification method based on the knowledge graph has a good effect of improving the defect classification.
It will be understood by those skilled in the art that the foregoing is only a preferred embodiment of the present invention, and is not intended to limit the invention, and although the invention has been described in detail with reference to the foregoing examples, it will be apparent to those skilled in the art that various changes in the form and details of the embodiments may be made and equivalents may be substituted for elements thereof. All modifications, equivalents and the like which come within the spirit and principle of the invention are intended to be included within the scope of the invention.
Claims (4)
1. A knowledge graph-based industrial product defect image classification method is characterized by comprising the following steps:
s1: respectively creating a defect map library X, a label library Y corresponding to each defect map and an additional attribute vector V corresponding to the defect type;
s2: and enhancing the defect map by utilizing an image enhancement technology to obtain an enhanced defect map library X'.
S3: constructing a defect image feature extraction network, and then continuously enhancing the defect image feature extraction network through S2Obtaining image batch X from the defect map library XBAnd its corresponding tag yB. Grouping x defective picturesBInputting the defect image feature extraction network to obtain image features of the image batch under different granularities, and normalizing to obtain normalized image features eB；
The defect image feature extraction network comprises a backbone network and a layered bilinear pooling network which are trained by Imagenet and used for extracting the VGG network features. After features u, v and w of the image under different granularities are extracted by a backbone network for extracting the VGG network features, cross-layer interaction of local characteristics is captured by utilizing a bilinear pooling network. The bilinear pooling network H has the following form:
H(u，v，w)＝PTconcat(UTu*VTv，UTu*WTw，VTv*WTw)
wherein P is a classification matrix, U, V and W are projection matrices of U, V and W respectively, and H (U, V and W) is the image characteristics of the image batches under different granularities.
S4: a plan embedding manner;
e obtained from S3BThe feature of each defect in (1) and the additional attribute vector V corresponding to the defect type in S1 are embedded into an undirected graph, thereby obtaining a graph batch g of batch image featuresB；
S5: constructing a graph convolution network for graph feature extraction, and dividing the graph batch g in S4BInputting the graph into the graph volume network to obtain a graph batch gBCharacteristic o ofB；
S6: for o in S5BFeature transformation is carried out to obtain transformed feature o'B。
S7: constructing a single-layer linear discriminant network, and then carrying out feature e in S3BO 'obtained from S6'BThe output of the discrimination network is a defect image batch xBCorresponding discrimination tag y'B；
S8: calculating defect image batch x from loss functionBTrue tag y ofBAnd discrimination tag y 'obtained in S7'BLoss between ld；
S9: simultaneous optimization of l with back-propagation algorithmdThe defect image feature extraction network, the graph convolution network and the discrimination network are carried out until the step ldConverging to finally obtain a classification model of the whole defect image;
s10: and performing the feature processing of the processes from S3 to S6 on the defect picture to be classified and the additional attribute vector corresponding to the defect type, and inputting the processed additional attribute vector into the classification model of the defect image obtained in S9, wherein the defect image feature extraction network, the graph convolution network and the single-layer linear discrimination network in S3-S6 are optimized networks, and finally the defect type to which the defect belongs is obtained.
2. The method for classifying defects of industrial products based on knowledge-graphs according to claim 1, wherein the enhancing the defect samples by image enhancing technique in S2 includes randomly turning over the negative samples, shifting, changing contrast, saturation, brightness and adding noise; meanwhile, when data is enhanced, the meaning of the defects is ensured not to change.
3. The method of classifying defects images of industrial products based on knowledge-maps according to claim 1, wherein said feature transformation of S6 is calculated by means of the following rules:
R(x)＝Conv1D(f(x))
wherein R is an inverse conversion function of the graph convolution characteristic, Conv1D is a one-dimensional convolution network, and f is a linear rectification function with a specific form as follows:
4. the knowledge-graph-based industrial product defect image classification method according to claim 1,
said S8Loss ldCalculated from the cross entropy loss, for which the discriminant class probability m and the defect class label n are entered, n ∈ Bj，
Wherein α is the weight of the class probability m, j is the number of defect classes, BjA vector set of defect class labels.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201911250467.0A CN111161213B (en) | 2019-12-09 | 2019-12-09 | Industrial product defect image classification method based on knowledge graph |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201911250467.0A CN111161213B (en) | 2019-12-09 | 2019-12-09 | Industrial product defect image classification method based on knowledge graph |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN111161213A true CN111161213A (en) | 2020-05-15 |
| CN111161213B CN111161213B (en) | 2022-03-11 |
Family
ID=70556603
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201911250467.0A Active CN111161213B (en) | 2019-12-09 | 2019-12-09 | Industrial product defect image classification method based on knowledge graph |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111161213B (en) |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112016601A (en) * | 2020-08-17 | 2020-12-01 | 华东师范大学 | Network model construction method based on knowledge graph enhanced small sample visual classification |
| CN113254650A (en) * | 2021-06-28 | 2021-08-13 | 明品云(北京)数据科技有限公司 | Knowledge graph-based assessment pushing method, system, equipment and medium |
| CN115660596A (en) * | 2022-11-03 | 2023-01-31 | 创启科技(广州)有限公司 | Data interaction method of mobile terminal and mobile terminal |
| CN116090559A (en) * | 2023-02-03 | 2023-05-09 | 深圳智现未来工业软件有限公司 | Method for generating knowledge points based on wafer map detection data |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107862239A (en) * | 2017-09-15 | 2018-03-30 | 广州唯品会研究院有限公司 | A kind of combination text carries out the method and its device of picture recognition with picture |
| CN108121829A (en) * | 2018-01-12 | 2018-06-05 | 扬州大学 | The domain knowledge collection of illustrative plates automated construction method of software-oriented defect |
| CN109325131A (en) * | 2018-09-27 | 2019-02-12 | 大连理工大学 | A kind of drug identification method based on biomedical knowledge map reasoning |
| US20190080245A1 (en) * | 2017-09-08 | 2019-03-14 | Niantic, Inc. | Methods and Systems for Generation of a Knowledge Graph of an Object |
| US20190095806A1 (en) * | 2017-09-28 | 2019-03-28 | Siemens Aktiengesellschaft | SGCNN: Structural Graph Convolutional Neural Network |
| CN110084296A (en) * | 2019-04-22 | 2019-08-02 | 中山大学 | A kind of figure expression learning framework and its multi-tag classification method based on certain semantic |
-
2019
- 2019-12-09 CN CN201911250467.0A patent/CN111161213B/en active Active
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20190080245A1 (en) * | 2017-09-08 | 2019-03-14 | Niantic, Inc. | Methods and Systems for Generation of a Knowledge Graph of an Object |
| CN107862239A (en) * | 2017-09-15 | 2018-03-30 | 广州唯品会研究院有限公司 | A kind of combination text carries out the method and its device of picture recognition with picture |
| US20190095806A1 (en) * | 2017-09-28 | 2019-03-28 | Siemens Aktiengesellschaft | SGCNN: Structural Graph Convolutional Neural Network |
| CN108121829A (en) * | 2018-01-12 | 2018-06-05 | 扬州大学 | The domain knowledge collection of illustrative plates automated construction method of software-oriented defect |
| CN109325131A (en) * | 2018-09-27 | 2019-02-12 | 大连理工大学 | A kind of drug identification method based on biomedical knowledge map reasoning |
| CN110084296A (en) * | 2019-04-22 | 2019-08-02 | 中山大学 | A kind of figure expression learning framework and its multi-tag classification method based on certain semantic |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112016601A (en) * | 2020-08-17 | 2020-12-01 | 华东师范大学 | Network model construction method based on knowledge graph enhanced small sample visual classification |
| CN112016601B (en) * | 2020-08-17 | 2022-08-05 | 华东师范大学 | Network model construction method based on knowledge graph enhanced small sample visual classification |
| CN113254650A (en) * | 2021-06-28 | 2021-08-13 | 明品云(北京)数据科技有限公司 | Knowledge graph-based assessment pushing method, system, equipment and medium |
| CN115660596A (en) * | 2022-11-03 | 2023-01-31 | 创启科技(广州)有限公司 | Data interaction method of mobile terminal and mobile terminal |
| CN115660596B (en) * | 2022-11-03 | 2023-07-04 | 创启科技(广州)有限公司 | Data interaction method of mobile terminal and mobile terminal |
| CN116090559A (en) * | 2023-02-03 | 2023-05-09 | 深圳智现未来工业软件有限公司 | Method for generating knowledge points based on wafer map detection data |
Also Published As
| Publication number | Publication date |
|---|---|
| CN111161213B (en) | 2022-03-11 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111161213B (en) | Industrial product defect image classification method based on knowledge graph | |
| CN109190752B (en) | Image semantic segmentation method based on global features and local features of deep learning | |
| CN108562589B (en) | Method for detecting surface defects of magnetic circuit material | |
| CN106875381B (en) | Mobile phone shell defect detection method based on deep learning | |
| CN108960245B (en) | Tire mold character detection and recognition method, device, equipment and storage medium | |
| Guan et al. | A steel surface defect recognition algorithm based on improved deep learning network model using feature visualization and quality evaluation | |
| CN107123088B (en) | A kind of method of automatic replacement photo background color | |
| CN112381788B (en) | Part surface defect increment detection method based on double-branch matching network | |
| CN111126115B (en) | Violent sorting behavior identification method and device | |
| CN110399840B (en) | Rapid lawn semantic segmentation and boundary detection method | |
| CN108921120B (en) | Cigarette identification method suitable for wide retail scene | |
| CN111860439A (en) | Unmanned aerial vehicle inspection image defect detection method, system and equipment | |
| CN112767369A (en) | Defect identification and detection method and device for small hardware and computer readable storage medium | |
| CN110880176B (en) | Semi-supervised industrial image defect segmentation method based on countermeasure generation network | |
| CN109948625A (en) | Definition of text images appraisal procedure and system, computer readable storage medium | |
| CN112037219A (en) | Metal surface defect detection method based on two-stage convolution neural network | |
| CN106203461B (en) | Image processing method and device | |
| CN110827304A (en) | Traditional Chinese medicine tongue image positioning method and system based on deep convolutional network and level set method | |
| CN115272204A (en) | Bearing surface scratch detection method based on machine vision | |
| CN114863464B (en) | Second-order identification method for PID drawing picture information | |
| CN115147418B (en) | Compression training method and device for defect detection model | |
| CN110852327A (en) | Image processing method, image processing device, electronic equipment and storage medium | |
| CN113298809A (en) | Composite material ultrasonic image defect detection method based on deep learning and superpixel segmentation | |
| CN115731400A (en) | X-ray image foreign matter detection method based on self-supervision learning | |
| CN115830359A (en) | Workpiece identification and counting method based on target detection and template matching in complex scene |
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 |