CN108074231B - Magnetic sheet surface defect detection method based on convolutional neural network - Google Patents
Magnetic sheet surface defect detection method based on convolutional neural network Download PDFInfo
- Publication number
- CN108074231B CN108074231B CN201711361710.7A CN201711361710A CN108074231B CN 108074231 B CN108074231 B CN 108074231B CN 201711361710 A CN201711361710 A CN 201711361710A CN 108074231 B CN108074231 B CN 108074231B
- Authority
- CN
- China
- Prior art keywords
- layer
- input
- neural network
- convolutional neural
- output
- 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.)
- Active
Links
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
- 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
- G06N3/045—Combinations of networks
-
- 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
- G06N3/048—Activation functions
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20081—Training; Learning
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20084—Artificial neural networks [ANN]
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Biophysics (AREA)
- Evolutionary Computation (AREA)
- Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- Biomedical Technology (AREA)
- Software Systems (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- Mathematical Physics (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Quality & Reliability (AREA)
- Image Analysis (AREA)
- Investigating Or Analyzing Materials By The Use Of Magnetic Means (AREA)
Abstract
The invention discloses a magnetic sheet surface defect detection method based on a convolutional neural network, which comprises the following steps: the method comprises the steps of firstly, obtaining an overlook image of a magnetic sheet to be detected and preprocessing the image, wherein the steps comprise graying, Hough circle transformation, size transformation, rotary cutting and the like; secondly, inputting the preprocessed image into a pre-trained convolutional neural network for defect detection, detecting whether the surface of the magnetic sheet has defects, and classifying the defects; the convolutional neural network comprises an input layer, a convolutional layer, a sampling layer and a full-connection layer, wherein the input layer, the convolutional layer and the sampling layer are used for extracting the characteristics of the image, and the extracted characteristics are subjected to defect classification by a Softmax classifier. Compared with the prior art, the method has high detection precision and better robustness.
Description
Technical Field
The invention belongs to a surface defect detection technology, and particularly relates to a magnetic sheet surface defect detection method based on a convolutional neural network.
Background
With the rapid development of electronic technology and computer technology, digital image processing technology has been widely used in many industries and fields, such as medical image processing and analysis, industrial control and detection automation, aerospace remote sensing and mapping, and the like, with the advantages of large information content, intuitive expression form, convenient transmission and storage, and the like. With the improvement of national economic level, people increasingly demand high-quality, high-precision and high-reliability products, and the problem is how to detect and judge whether the products manufactured in large quantities reach performance indexes.
The traditional detection method is implemented manually, the labor intensity of workers in manual detection is high, the detection is limited by various factors such as the mental state of the workers, the detection proficiency level, the experience accumulation level and the working environment, the detection efficiency is low, the speed is low, and the consistency of devices is difficult to guarantee. In the detection process, due to the fatigue of workers, wrong picking and missed detection are inevitably generated, and the outflow of unqualified products not only can bring economic loss to a factory, but also can bring potential safety hazards to users more seriously. Therefore, how to detect the surface defects of the parts quickly, efficiently and accurately becomes an urgent problem to be solved by the manufacturing industry. The improvement of computer technology enables the digital image processing technology to be widely applied, and the surface defect detection system based on machine vision has high application value.
At present, the surface defect detection method based on machine vision mainly comprises two methods: one is a feature processing method based on an image processing algorithm, and the other is a method using a convolutional neural network in combination with an image processing algorithm. The characteristic processing method based on the image processing algorithm comprises the steps of respectively calculating the average graying value and the variance of the graying value of an image through graying processing of the image, comparing the average graying value and the variance of the graying value with a preset standard value, and judging whether the difference value between the average graying value and the standard value exceeds a threshold value or not to judge whether surface defects exist or not. The method usually needs to preset information such as exposure time of a camera, brightness of ambient light, positions acquired by magnetic sheets and the like, is greatly influenced by external conditions, has poor robustness of the system and solves a single defect problem. And the method of combining the convolutional neural network and the image processing algorithm, wherein the convolutional network is a multi-layer perceptron specially designed for identifying two-dimensional shapes, and the network structure has high invariance to translation, scaling, inclination or other forms of deformation. And the complex characteristic extraction and data reconstruction process in the traditional identification algorithm is avoided. The recognition speed of the defect features is greatly improved, but the method needs a large number of samples for training to reach a high recognition rate, and the recognition rate is reduced due to too few training samples.
The surface detection technology mainly relates to the steps of image acquisition, image digital processing, construction and training of a convolutional neural network, classification of image defects and the like. The digital processing of images and the construction of convolutional neural networks are two key problems in the field of surface detection.
Disclosure of Invention
The invention aims to overcome the defects of the prior art and provides a magnetic sheet surface defect detection method based on a convolutional neural network, so that the detection of various defects on the surface of a magnetic sheet can be rapidly and accurately realized.
In order to achieve the purpose, the technical scheme of the invention is realized as follows:
a magnetic sheet surface defect detection method based on a convolutional neural network comprises the following steps:
(1) acquiring a top view image of a magnetic sheet to be detected and preprocessing the image;
(2) inputting the preprocessed image into a pre-trained convolutional neural network for defect detection, detecting whether the surface of the magnetic sheet has defects, and classifying the defects; the convolutional neural network comprises an input layer, a convolutional layer, a sampling layer and a full-connection layer, wherein the input layer, the convolutional layer and the sampling layer are used for extracting the characteristics of the image, and the extracted characteristics are subjected to defect classification by a Softmax classifier.
The preprocessing of the magnetic sheet image specifically comprises the following steps:
the method comprises the steps of firstly, carrying out gray processing on a magnetic sheet image, carrying out Hough circle transformation on the gray image, detecting the outline of a magnetic sheet, and cutting a minimum circumscribed square of a circle according to the circle center and the circle radius;
secondly, the cut square images are used as templates, and the residual images are subjected to batch template matching processing, so that all the images are the minimum circumscribed circle size of the outer contour of the magnetic sheet;
thirdly, converting the sizes of all the images obtained in the second step to set sizes;
fourthly, rotationally dividing the size-transformed image by using an image processing method, and dividing a large image into a plurality of small images;
the training method of the convolutional neural network is as follows:
(a) and collecting data samples. Collecting a large number of magnetic sheet images including non-defective magnetic sheets and defective magnetic sheets;
(b) the data set is augmented. Preprocessing the acquired magnetic sheet, rotationally dividing the image, cutting a magnetic sheet image into a plurality of images, manually marking the images into 4 types including a defective sand eye type, a defective block type, a defective crack type and a non-defective type, and dividing the obtained data samples into two parts according to a certain proportion, wherein one part is used as a training set and the other part is used as a test set;
(c) establishing a convolutional neural network;
(d) inputting the training set sample data into a convolutional neural network, training the convolutional neural network, and evaluating the training effect of the convolutional neural network by using the test set sample data;
the structure of the convolutional neural network is as follows:
an input layer: the first layer is an input layer, and the input size is the picture size 43 x 43;
layer C1: the output of the input layer is 43 × 43 matrix, the input of the C1 convolution layer is the output of the input layer, the input is convolved with 64 convolution kernels with 3 × 3 step 1, the excitation function is RELU, 64 feature maps are extracted in total, and the output is 41 × 64 matrix;
layer S2: the input of the S2 pooling layer is the output of the C1 convolution layer, the maximum value sampling processing is carried out on 64 input maps by using convolution kernels with 2 x 2 stepping to 2, and the output is a matrix of 20 x 64;
layer C3: the input of the layer C3 is the output of the layer S2, the convolution operation is carried out on 64 input maps by using 128 convolution kernels with 3 × 3 stepping as 1, the excitation function is RELU, 128 characteristic maps are extracted in total, and the output is a matrix of 18 × 128;
layer S4: the input of the S4 layer is the output of the C3 layer, the maximum value sampling processing is carried out on the 128 input maps by using a convolution kernel with 2 x 2 stepping to 2, and the output is a matrix of 9 x 128;
layer C5: the input of the layer C5 is the output of the layer S4, the convolution operation is carried out on the 128 input maps by using 256 convolution kernels with 3 × 3 stepping as 1, the excitation function is RELU, 256 feature maps are extracted in total, and the output is 7 × 256;
f6 full connection layer: the full-connection layer 1 carries out weighted calculation on the extracted features to obtain a 1-dimensional feature vector;
f7 full connection layer 2: the full-connection layer 2 performs weighted calculation on the feature vectors output by the full-connection layer 1 to obtain 1-dimensional feature vectors with more concentrated features;
softmax classifier: classifying the feature vectors at the output end of the full connection layer 2;
compared with the prior art, the invention has the following remarkable advantages: (1) the invention preprocesses the image to be detected, which is equivalent to increase the number of training samples and improves the problem of poor neural network training caused by the lack of the training samples. (2) In the traditional defect detection method based on the digital image, the identification algorithm comprises complex characteristic extraction and data reconstruction processes, and each defect needs a separate identification algorithm. (3) Compared with the traditional detection method based on image processing, the detection method based on the convolutional neural network is higher in robustness, the image is influenced by external factors such as light change in the image acquisition stage, the traditional method is likely to not extract features, but the method of the convolutional neural network can intentionally add a part of pictures influenced by the external factors when the neural network is trained, the detection accuracy of the convolutional neural network in the aspect is enhanced, and the detection method is more in line with the requirements of industrial fields.
Drawings
FIG. 1 is a flow chart of the method of the present invention;
FIG. 2 is a flow chart of the pre-treatment of the surface image of the magnetic sheet to be detected according to the present invention;
FIG. 3 is an image of the surface of a magnetic sheet to be inspected after a pretreatment;
FIG. 4 is a schematic diagram of a network of convolutional neural networks of the present invention;
Detailed Description
The technical solution in the embodiments of the present invention is clearly and completely described below with reference to the drawings in the embodiments of the present invention.
Referring to fig. 1 to 4, a magnetic sheet defect detection method based on a convolutional neural network includes the following processes: training material was first collected, with 1000 disks containing defects and 500 disks without defects. Acquiring top views of two surfaces of a magnetic sheet in a standard environment, performing graying processing on the magnetic sheet image through an image preprocessing process shown in fig. 2, performing Hough circle transformation on the grayscale image, detecting the outline of the magnetic sheet, and cutting the minimum circumscribed square of the circle according to the circle center and the circle radius; and then, the cut square images are used as templates, the residual images are subjected to batch template matching processing, all the images are the minimum circumscribed circle size of the outer contour of the magnetic sheet, all the images are subjected to size conversion to a set size, an image processing method is used for rotary segmentation, and a large image is segmented into a plurality of small images. And manually calibrating the obtained image, and balancing the defect proportion to obtain 1000 trachoma classes, 1000 defect block classes, 1000 scratch classes and 1000 defect-free picture materials, wherein 700 pictures are selected to be put into a training set, and 300 pictures are put into a testing set. And putting the training set and the test set into a convolutional neural network under a Tensorflow architecture in sequence, wherein the trained convolutional neural network is a model for detecting the defects of the magnetic sheets.
The structure of the convolutional neural network is as follows:
an input layer: the first layer is an input layer, and the input size is the picture size 43 x 43;
layer C1: the output of the input layer is 43 × 43 matrix, the input of the C1 convolution layer is the output of the input layer, the input is convolved with 64 convolution kernels with 3 × 3 step 1, the excitation function is RELU, 64 feature maps are extracted in total, and the output is 41 × 64 matrix;
layer S2: the input of the S2 pooling layer is the output of the C1 convolution layer, the maximum value sampling processing is carried out on 64 input maps by using convolution kernels with 2 x 2 stepping to 2, and the output is a matrix of 20 x 64;
layer C3: the input of the layer C3 is the output of the layer S2, the convolution operation is carried out on 64 input maps by using 128 convolution kernels with 3 × 3 stepping as 1, the excitation function is RELU, 128 characteristic maps are extracted in total, and the output is a matrix of 18 × 128;
layer S4: the input of the S4 layer is the output of the C3 layer, the maximum value sampling processing is carried out on the 128 input maps by using a convolution kernel with 2 x 2 stepping to 2, and the output is a matrix of 9 x 128;
layer C5: the input of the layer C5 is the output of the layer S4, the convolution operation is carried out on the 128 input maps by using 256 convolution kernels with 3 × 3 stepping as 1, the excitation function is RELU, 256 feature maps are extracted in total, and the output is 7 × 256;
f6 full connection layer: the full-connection layer 1 carries out weighted calculation on the extracted features to obtain a 1-dimensional feature vector;
f7 full connection layer 2: the full-connection layer 2 performs weighted calculation on the feature vectors output by the full-connection layer 1 to obtain 1-dimensional feature vectors with more concentrated features;
softmax classifier: classifying the feature vectors at the output end of the full connection layer 2;
the output layer includes 5 cells, representing 5 cases respectively: (1) crack type (2) lack block type (3), sand eye type (4) lack defect type (5) and other types.
The embodiments described in this specification are merely illustrative of implementations of the inventive concept and the scope of the present invention should not be considered limited to the specific forms set forth in the embodiments but rather by the equivalents thereof as may occur to those skilled in the art upon consideration of the present inventive concept.
Claims (1)
1. A magnetic sheet surface defect detection method based on a convolutional neural network comprises the following steps:
(1) acquiring a top view image of a magnetic sheet to be detected and preprocessing the image;
the method specifically comprises the following steps:
the method comprises the steps of firstly, carrying out gray processing on a magnetic sheet image, carrying out Hough circle transformation on the gray image, detecting the outline of a magnetic sheet, and cutting a minimum circumscribed square of a circle according to the circle center and the circle radius;
secondly, the cut square images are used as templates, and the residual images are subjected to batch template matching processing, so that all the images are the minimum circumscribed circle size of the outer contour of the magnetic sheet;
thirdly, converting the sizes of all the images obtained in the second step to set sizes;
fourthly, rotationally dividing the size-transformed image by using an image processing method, and dividing a large image into a plurality of small images;
(2) inputting the preprocessed image into a pre-trained convolutional neural network for defect detection, detecting whether the surface of the magnetic sheet has defects, and classifying the defects; a convolutional neural network in which an input layer, a convolutional layer, a sampling layer,
the full connection layer extracts the features of the image, and the extracted features are subjected to defect classification by a Softmax classifier; the training method of the convolutional neural network comprises the following steps:
(a) collecting data samples, namely collecting a large number of magnetic sheet images including non-defective magnetic sheets and defective magnetic sheets;
(b) expanding a data set, preprocessing the acquired magnetic sheet images, rotationally dividing the images, cutting a pair of magnetic sheet images into a plurality of images, manually marking the images into 4 types including a defective sand eye type, a defective block type, a defective crack type and a defect-free type, dividing the obtained data samples into two parts according to a certain proportion, wherein one part is used as a training set, and the other part is used as a testing set;
(c) establishing a convolutional neural network;
(d) inputting the training set sample data into a convolutional neural network, training the convolutional neural network, and evaluating the training effect of the convolutional neural network by using the test set sample data; the convolutional neural network is as follows:
an input layer: the first layer is an input layer, and the input size is the picture size 43 x 43;
layer C1: the output of the input layer is 43 × 43 matrix, the input of the C1 convolution layer is the output of the input layer, the input is convolved with 64 convolution kernels with 3 × 3 step 1, the excitation function is RELU, 64 feature maps are extracted in total, and the output is 41 × 64 matrix;
layer S2: the input of the S2 pooling layer is the output of the C1 convolution layer, the maximum value sampling processing is carried out on 64 input maps by using convolution kernels with 2 x 2 stepping to 2, and the output is a matrix of 20 x 64;
layer C3: the input of the layer C3 is the output of the layer S2, the convolution operation is carried out on 64 input maps by using 128 convolution kernels with 3 × 3 stepping as 1, the excitation function is RELU, 128 characteristic maps are extracted in total, and the output is a matrix of 18 × 128;
layer S4: the input of the S4 layer is the output of the C3 layer, the maximum value sampling processing is carried out on the 128 input maps by using a convolution kernel with 2 x 2 stepping to 2, and the output is a matrix of 9 x 128;
layer C5: the input of the layer C5 is the output of the layer S4, the convolution operation is carried out on the 128 input maps by using 256 convolution kernels with 3 × 3 stepping as 1, the excitation function is RELU, 256 feature maps are extracted in total, and the output is 7 × 256;
f6 full connection layer: the full-connection layer 1 carries out weighted calculation on the extracted features to obtain a 1-dimensional feature vector;
f7 full connection layer 2: the full-connection layer 2 performs weighted calculation on the feature vectors output by the full-connection layer 1 to obtain 1-dimensional feature vectors with more concentrated features;
softmax classifier: and classifying the feature vectors at the output end of the full connection layer 2.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201711361710.7A CN108074231B (en) | 2017-12-18 | 2017-12-18 | Magnetic sheet surface defect detection method based on convolutional neural network |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201711361710.7A CN108074231B (en) | 2017-12-18 | 2017-12-18 | Magnetic sheet surface defect detection method based on convolutional neural network |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN108074231A CN108074231A (en) | 2018-05-25 |
| CN108074231B true CN108074231B (en) | 2020-04-21 |
Family
ID=62158463
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201711361710.7A Active CN108074231B (en) | 2017-12-18 | 2017-12-18 | Magnetic sheet surface defect detection method based on convolutional neural network |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108074231B (en) |
Families Citing this family (20)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108685158B (en) * | 2018-05-28 | 2021-11-09 | 杭州安脉盛智能技术有限公司 | Intelligent filament cutter maintenance method and system based on automatic optical detection technology |
| CN108985337A (en) * | 2018-06-20 | 2018-12-11 | 中科院广州电子技术有限公司 | A kind of product surface scratch detection method based on picture depth study |
| CN108960413A (en) * | 2018-07-11 | 2018-12-07 | 天津工业大学 | A kind of depth convolutional neural networks method applied to screw surface defects detection |
| CN109064454A (en) * | 2018-07-12 | 2018-12-21 | 上海蝶鱼智能科技有限公司 | Product defects detection method and system |
| CN109300109B (en) * | 2018-08-23 | 2020-10-09 | 苏州富鑫林光电科技有限公司 | Machine learning industrial intelligent image segmentation and recognition system |
| CN109239075B (en) * | 2018-08-27 | 2021-11-30 | 北京百度网讯科技有限公司 | Battery detection method and device |
| CN109376773A (en) * | 2018-09-30 | 2019-02-22 | 福州大学 | Crack detecting method based on deep learning |
| CN109800708A (en) * | 2018-12-13 | 2019-05-24 | 程琳 | Visit image lesion intelligent identification Method in aero-engine hole based on deep learning |
| CN109685968A (en) * | 2018-12-15 | 2019-04-26 | 西安建筑科技大学 | A kind of the identification model building and recognition methods of the banknote image defect based on convolutional neural networks |
| CN111292345A (en) * | 2019-05-17 | 2020-06-16 | 展讯通信（上海）有限公司 | Purple border detection method and device, electronic equipment and storage medium |
| CN110243934A (en) * | 2019-05-30 | 2019-09-17 | 中国计量大学上虞高等研究院有限公司 | A kind of ultrasonic weld seam detection method based on wavelet convolution neural network |
| CN110288013A (en) * | 2019-06-20 | 2019-09-27 | 杭州电子科技大学 | A kind of defective labels recognition methods based on block segmentation and the multiple twin convolutional neural networks of input |
| CN110929773A (en) * | 2019-11-18 | 2020-03-27 | 上海集成电路研发中心有限公司 | Chip surface defect classification device and method based on deep learning |
| CN111062391B (en) * | 2019-12-25 | 2023-09-19 | 创新奇智(青岛)科技有限公司 | Magnetic sheet initial positioning method |
| CN112419291B (en) * | 2020-11-30 | 2024-06-04 | 佛山职业技术学院 | Training method of bottle embryo defect detection model, storage medium and terminal equipment |
| CN112819824B (en) * | 2021-03-03 | 2024-04-26 | 科为升视觉技术（苏州）有限公司 | Neural network for vision inspection system, inspection system and method including the same |
| CN113516650B (en) * | 2021-07-30 | 2023-08-25 | 深圳康微视觉技术有限公司 | Circuit board hole plugging defect detection method and device based on deep learning |
| CN113610795B (en) * | 2021-08-02 | 2023-09-29 | 沈阳航空航天大学 | Method and system for detecting surface defects of combustible cartridge |
| CN113793332B (en) * | 2021-11-15 | 2022-02-08 | 山东德普检测技术有限公司 | Experimental instrument defect identification and classification method and system |
| CN116091505B (en) * | 2023-04-11 | 2023-06-30 | 青岛芯康半导体科技有限公司 | Automatic defect detection and classification method and system for sapphire substrate |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106127780A (en) * | 2016-06-28 | 2016-11-16 | 华南理工大学 | A kind of curved surface defect automatic testing method and device thereof |
| CN107169556A (en) * | 2017-05-15 | 2017-09-15 | 电子科技大学 | stem cell automatic counting method based on deep learning |
| CN107316295A (en) * | 2017-07-02 | 2017-11-03 | 苏州大学 | A kind of fabric defects detection method based on deep neural network |
| CN107392896A (en) * | 2017-07-14 | 2017-11-24 | 佛山市南海区广工大数控装备协同创新研究院 | A kind of Wood Defects Testing method and system based on deep learning |
| CN107481231A (en) * | 2017-08-17 | 2017-12-15 | 广东工业大学 | A kind of handware defect classifying identification method based on depth convolutional neural networks |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US10186026B2 (en) * | 2015-11-17 | 2019-01-22 | Kla-Tencor Corp. | Single image detection |
-
2017
- 2017-12-18 CN CN201711361710.7A patent/CN108074231B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106127780A (en) * | 2016-06-28 | 2016-11-16 | 华南理工大学 | A kind of curved surface defect automatic testing method and device thereof |
| CN107169556A (en) * | 2017-05-15 | 2017-09-15 | 电子科技大学 | stem cell automatic counting method based on deep learning |
| CN107316295A (en) * | 2017-07-02 | 2017-11-03 | 苏州大学 | A kind of fabric defects detection method based on deep neural network |
| CN107392896A (en) * | 2017-07-14 | 2017-11-24 | 佛山市南海区广工大数控装备协同创新研究院 | A kind of Wood Defects Testing method and system based on deep learning |
| CN107481231A (en) * | 2017-08-17 | 2017-12-15 | 广东工业大学 | A kind of handware defect classifying identification method based on depth convolutional neural networks |
Non-Patent Citations (1)
| Title |
|---|
| R. Liu etc.Region-convolutional neural network for detecting capsule surface defects.《Boletin Tecnico/Technical Bulletin》.2017,全文. * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN108074231A (en) | 2018-05-25 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108074231B (en) | Magnetic sheet surface defect detection method based on convolutional neural network | |
| CN108562589B (en) | Method for detecting surface defects of magnetic circuit material | |
| CN111179251B (en) | Defect detection system and method based on twin neural network and by utilizing template comparison | |
| CN108305243B (en) | Magnetic shoe surface defect detection method based on deep learning | |
| CN112037219B (en) | Metal surface defect detection method based on two-stage convolutional neural network | |
| CN111582294B (en) | Method for constructing convolutional neural network model for surface defect detection and application thereof | |
| CN113643268B (en) | Industrial product defect quality inspection method and device based on deep learning and storage medium | |
| CN111667455A (en) | AI detection method for various defects of brush | |
| CN113298757A (en) | Metal surface defect detection method based on U-NET convolutional neural network | |
| CN106290392A (en) | A kind of little micro-bearing surface pitting defects online test method and system thereof | |
| CN106355579A (en) | Defect detecting method of cigarette carton surface wrinkles | |
| CN111951232A (en) | Metal powder injection molding appearance defect detection method and system | |
| CN111179250A (en) | Industrial product defect detection system based on multitask learning | |
| WO2024002187A1 (en) | Defect detection method, defect detection device, and storage medium | |
| CN113177924A (en) | Industrial production line product flaw detection method | |
| CN113506239B (en) | Strip steel surface defect detection method based on cross-stage local network | |
| CN116071315A (en) | Product visual defect detection method and system based on machine vision | |
| CN117434082A (en) | Quality detection method and system for silica gel product production line | |
| CN110232703B (en) | Moving object recognition device and method based on color and texture information | |
| CN114841957A (en) | Steel plate surface defect detection method based on deep learning | |
| CN113989196A (en) | Vision-based earphone silica gel gasket appearance defect detection method | |
| Zhang et al. | Design of tire damage image recognition system based on deep learning | |
| CN112396580A (en) | Circular part defect detection method | |
| Bhutta et al. | Smart-inspect: micro scale localization and classification of smartphone glass defects for industrial automation | |
| CN107122783B (en) | Method for quickly identifying assembly connector based on angular point detection |
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 |