CN109472769A - A kind of bad image defect detection method and system - Google Patents
A kind of bad image defect detection method and system Download PDFInfo
- Publication number
- CN109472769A CN109472769A CN201811123897.1A CN201811123897A CN109472769A CN 109472769 A CN109472769 A CN 109472769A CN 201811123897 A CN201811123897 A CN 201811123897A CN 109472769 A CN109472769 A CN 109472769A
- Authority
- CN
- China
- Prior art keywords
- image
- defect
- detection
- bad
- training
- 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
- G06T7/001—Industrial image inspection using an image reference approach
-
- 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
- G06F18/2413—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on distances to training or reference patterns
-
- 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
- G06F18/2413—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on distances to training or reference patterns
- G06F18/24147—Distances to closest patterns, e.g. nearest neighbour classification
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/70—Determining position or orientation of objects or cameras
- G06T7/73—Determining position or orientation of objects or cameras using feature-based methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10004—Still image; Photographic image
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20081—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]
-
- 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
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- General Physics & Mathematics (AREA)
- Physics & Mathematics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- General Engineering & Computer Science (AREA)
- Quality & Reliability (AREA)
- Image Analysis (AREA)
Abstract
The present invention discloses a kind of bad image defect detection method and system, pass through deep learning method, depth convolutional neural networks model is constructed, deep layer logical relation, part and local, part and whole defect relationship in bad image between defect factors are analyzed, realization defect is classified;The present invention can effectively realize defective locations detection and defect class prediction, and precision is high and classification is accurate, more preferable to the detection effect of Small object;Can optimizing detection precision, reduce missing inspection.
Description
Technical field
The invention belongs to image deflects detection technique fields, more particularly to a kind of bad image defect detection method and are
System.
Background technique
The product produced in the process of production and processing can have many bad defects, can usually pass through the side of image recognition
The detection of formula progress product defects.But this defect classification is more, the characteristics of image multiplicity of defect of the same race, without notable feature, portion
Divide the image appearance of defect unobvious;The mode detection image defect of existing Feature Engineering, is difficult to be accurately detected bad image
In various defects.
Traditional images detection method, if new feature and the feature set set have a little discrepancy, conventional machines
Habit can not accurately detect defect, and overall precision is caused to decline.It is big to realize that automatic detection enters difficulty, need it is experienced specially
Family manually sets defect characteristic, and according to the continuous variation synchronized update defect characteristic of production line and technique.
Summary of the invention
It to solve the above-mentioned problems, can be effective the invention proposes a kind of bad image defect detection method and system
Realize defective locations detection and defect class prediction, precision high-class is accurate, more preferable to the detection effect of Small object;It can optimize
Detection accuracy reduces missing inspection.
In order to achieve the above objectives, the technical solution adopted by the present invention is that: a kind of bad image defect detection method, including step
It is rapid:
S1 uploads defect image data and normal image data;
S2 is marked the rejected region on defect image using mark tool;
Image data is divided into two class of training sample and test sample by S3;
S4 is trained according to different defect levels using the corresponding neural network for calculating rank, training deep learning mould
Type;
S5 carries out defect location by template matching method;
S6 is trained test sample by deep learning model, detects bad image deflects pixel region and image
The specific location that middle defect occurs, predicts bad image deflects classification;
S7, plus test sample again repetitive exercise deep learning model on the basis of original training sample.
Further, the neural network is divided into miniature neural net according to different calculating ranks in the step S4
Network, plain edition neural network, large-scale neural network and ultra-large type neural network.
Further, in the step S5, the template matching method comprising steps of
Template Location: the maximum cycle temper figure in input picture is searched;
Template matching: the consistency by analyzing image and all templates determines whether image is defect image;Described
Template comparison is carried out in threshold range, is then defect image if more than threshold range, is normal picture if being less than threshold value.
Further, the step of Template Location, includes:
Firstly, carrying out characteristic point detection to image;
Then, characteristic point is analyzed, the characteristic point with similar characteristics is grouped;
Finally, obtaining the periodical subgraph in image by the topological relation of characteristic point in analysis group.
Further, establishing Mask RCNN convolution net in the neural network training process in the step S6
Network, comprising steps of
In conjunction with the dividing method of full convolutional network FCN and RCNN target analyte detection, image is detected, is exported in image
The classification and detection block of target;Pixel segmentation is carried out based on classification and detection block;
Corresponding multilayer RPN Area generation net is constructed on the convolutional layer in three layers of different feeling domain based on image pyramid thought
Network.
Further, the algorithm based on Mask RCNN convolutional network deep learning carries out in the step S6
Detection:
By full convolutional network FCN, converts out based on the original image after pixel detection, detect bad image deflects picture
Plain region;
Based on homing method, the specific location of bad defect appearance in detection image;
Bad image deflects classification is predicted using cross entropy loss function using softmax classifier.
Further, in the step S7, to deep learning model iterative increment learning method, comprising steps of
Training process: prepare calibration sample;It establishes and initializes network architecture;Training pattern passes through loss function
Callback model parameter;
Training end mark: when loss function no longer changes, model convergence, training stops.
On the other hand, the present invention also provides a kind of bad image deflects detection systems, including image collection module, sample
Division module, defect location module, model management module and defects detection module;
Image collection module receives and uploads defect image data and normal image data;
The image data of upload is divided into two class of training sample and test sample by sample division module;
Defect location module carries out defect location to image data;
Model management module, foundation, training and the optimization of implementation model;According to different defect levels, calculated using corresponding
The neural network of rank is trained, training deep learning model;Test sample weight is added on the basis of original training sample
New repetitive exercise deep learning model;
Defects detection module is trained test sample by template comparison and deep learning model, detects not plan deliberately
As the specific location that defect occurs in defect pixel region and image, bad image deflects classification is predicted.
Using the technical program the utility model has the advantages that
The present invention constructs depth convolutional neural networks model by deep learning method, obtains bad image by study
Deep layer logical relation, part and part, part and whole defect relationship between middle defect factors, realize defect classification;Energy
Automatic detection level is enough realized and improved, by deep learning algorithm and product image visualization interface, can both realize model
Training, model application judge image with the presence or absence of defect and detect defective locations and prediction defect classification;The present invention uses needle
To the peculiar algorithm framework and algorithm of industry, it is ensured that industrial detection accuracy;For emerging defect, covered using incremental learning
New defect is covered, optimizing detection precision reduces missing inspection;The present invention can substitute existing manual detection mode, improve production efficiency,
To improve production capacity;
The present invention passes sequentially through Template Location mode and neural metwork training mode combines, and improves calculating speed;It can
The defects detection for realizing different scale, improves the valve deck face of defects detection, it is ensured that industrial detection accuracy reduces missing inspection;
The present invention uses the Pixel-level algorithm of target detection based on deep learning, so that the defects detection precision of bad image
High, classification is accurately, more preferable especially for the detection effect of Small object defect;
The present invention is based on the detection methods of machine learning can be subtracted by model repetitive exercise mode, optimizing detection precision
Few missing inspection;By way of substitution model or model retraining, new detection demand is identified;Simultaneously using based on image procossing
Template matching algorithm verifies the accuracy and missing inspection situation of detection algorithm, the upgrading of Optimized model iteration.
Detailed description of the invention
Fig. 1 is a kind of flow diagram of bad image defect detection method of the invention;
Fig. 2 is the flow diagram of defect location in the embodiment of the present invention;
Fig. 3 is a kind of structural schematic diagram of bad image deflects detection system in the embodiment of the present invention.
Specific embodiment
To make the objectives, technical solutions, and advantages of the present invention clearer, the present invention is made into one with reference to the accompanying drawing
Step illustrates.
In the present embodiment, shown in Figure 1, the invention proposes a kind of bad image defect detection methods, including step
It is rapid:
S1 uploads defect image data and normal image data;
S2 is marked the rejected region on defect image using mark tool;
Image data is divided into two class of training sample and test sample by S3;
S4 is trained according to different defect levels using the corresponding neural network for calculating rank, training deep learning mould
Type;
S5 carries out defect location by template matching method;
S6 is trained test sample by deep learning model, detects bad image deflects pixel region and image
The specific location that middle defect occurs, predicts bad image deflects classification;
S7, plus test sample again repetitive exercise deep learning model on the basis of original training sample.
As the prioritization scheme of above-described embodiment, in the step S4, the neural network is according to different calculating ranks
It is divided into miniature neural network, plain edition neural network, large-scale neural network and ultra-large type neural network.
As the prioritization scheme of above-described embodiment, in the step S5, the template matching method comprising steps of
Template Location: the maximum cycle temper figure in input picture is searched;
Template matching: the consistency by analyzing image and all templates determines whether image is defect image;Described
Template comparison is carried out in threshold range, is then defect image if more than threshold range, is normal picture if being less than threshold value.
The step of Template Location includes:
Firstly, carrying out characteristic point detection to image;
Then, characteristic point is analyzed, the characteristic point with similar characteristics is grouped;
Finally, obtaining the periodical subgraph in image by the topological relation of characteristic point in analysis group.
It is established in the neural network training process in the step S6 as the prioritization scheme of above-described embodiment
Mask RCNN convolutional network, comprising steps of
In conjunction with the dividing method of full convolutional network FCN and RCNN target analyte detection, image is detected, is exported in image
The classification and detection block of target;Pixel segmentation is carried out based on classification and detection block;
Corresponding multilayer RPN Area generation net is constructed on the convolutional layer in three layers of different feeling domain based on image pyramid thought
Network.
In the step S6, the algorithm based on Mask RCNN convolutional network deep learning is detected:
By full convolutional network FCN, converts out based on the original image after pixel detection, detect bad image deflects picture
Plain region;
Based on homing method, the specific location of bad defect appearance in detection image;
Bad image deflects classification is predicted using cross entropy loss function using softmax classifier.
As the prioritization scheme of above-described embodiment, in the step S7, to deep learning model iterative increment study side
Method, comprising steps of
Training process: prepare calibration sample;It establishes and initializes network architecture;Training pattern passes through loss function
Callback model parameter;
Training end mark: when loss function no longer changes, model convergence, training stops.
For the realization for cooperating the method for the present invention, it is based on identical inventive concept, as shown in figure 3, the present invention also provides one
The bad image deflects detection system of kind, including image collection module, sample division module, defect location module, model management mould
Block and defects detection module;
Image collection module receives and uploads defect image data and normal image data;
The image data of upload is divided into two class of training sample and test sample by sample division module;
Defect location module carries out defect location to image data;
Model management module, foundation, training and the optimization of implementation model;According to different defect levels, calculated using corresponding
The neural network of rank is trained, training deep learning model;Test sample weight is added on the basis of original training sample
New repetitive exercise deep learning model;
Defects detection module is trained test sample by template comparison and deep learning model, detects not plan deliberately
As the specific location that defect occurs in defect pixel region and image, bad image deflects classification is predicted.
The above shows and describes the basic principles and main features of the present invention and the advantages of the present invention.The technology of the industry
Personnel are it should be appreciated that the present invention is not limited to the above embodiments, and the above embodiments and description only describe this
The principle of invention, without departing from the spirit and scope of the present invention, various changes and improvements may be made to the invention, these changes
Change and improvement all fall within the protetion scope of the claimed invention.The claimed scope of the invention by appended claims and its
Equivalent thereof.
Claims (8)
1. a kind of bad image defect detection method, which is characterized in that comprising steps of
S1 uploads defect image data and normal image data;
S2 is marked the rejected region on defect image using mark tool;
Image data is divided into two class of training sample and test sample by S3;
S4 is trained according to different defect levels using the corresponding neural network for calculating rank, training deep learning model;
S5 carries out defect location by template matching method;
S6 is trained test sample by deep learning model, detects in bad image deflects pixel region and image and lacks
Trap out existing specific location, predicts bad image deflects classification;
S7, plus test sample again repetitive exercise deep learning model on the basis of original training sample.
2. a kind of bad image defect detection method according to claim 1, which is characterized in that in the step S4,
The neural network is divided into miniature neural network, plain edition neural network, large-scale neural network according to different calculating ranks and surpasses
Large-scale neural network.
3. a kind of bad image defect detection method according to claim 2, which is characterized in that in the step S5,
The template matching method comprising steps of
Template Location: the maximum cycle temper figure in input picture is searched;
Template matching: the consistency by analyzing image and all templates determines whether image is defect image;In the threshold value
Template comparison is carried out in range, is then defect image if more than threshold range, is normal picture if being less than threshold value.
4. a kind of bad image defect detection method according to claim 3, which is characterized in that the step of the Template Location
Suddenly include:
Firstly, carrying out characteristic point detection to image;
Then, characteristic point is analyzed, the characteristic point with similar characteristics is grouped;
Finally, obtaining the periodical subgraph in image by the topological relation of characteristic point in analysis group.
5. a kind of bad image defect detection method according to claim 4, which is characterized in that in the step S6,
Mask RCNN convolutional network is established in the neural network training process, comprising steps of
In conjunction with the dividing method of full convolutional network FCN and RCNN target analyte detection, image is detected, exports target in image
Classification and detection block;Pixel segmentation is carried out based on classification and detection block;
Corresponding multilayer RPN Area generation network is constructed on the convolutional layer in three layers of different feeling domain based on image pyramid thought.
6. a kind of bad image defect detection method according to claim 5, which is characterized in that in the step S6,
The algorithm based on Mask RCNN convolutional network deep learning is detected:
By full convolutional network FCN, converts out based on the original image after pixel detection, detect bad image deflects pixel region
Domain;
Based on homing method, the specific location of bad defect appearance in detection image;
Bad image deflects classification is predicted using cross entropy loss function using softmax classifier.
7. a kind of bad image defect detection method according to claim 7, which is characterized in that in the step S7,
To deep learning model iterative increment learning method, comprising steps of
Training process: prepare calibration sample;It establishes and initializes network architecture;Training pattern is adjusted back by loss function
Model parameter;
Training end mark: when loss function no longer changes, model convergence, training stops.
8. a kind of bad image deflects detection system, which is characterized in that including image collection module, sample division module, defect
Locating module, model management module and defects detection module;
Image collection module receives and uploads defect image data and normal image data;
The image data of upload is divided into two class of training sample and test sample by sample division module;
Defect location module carries out defect location to image data;
Model management module, foundation, training and the optimization of implementation model;According to different defect levels, rank is calculated using corresponding
Neural network be trained, training deep learning model;It changes again on the basis of original training sample plus test sample
Generation training deep learning model;
Defects detection module is trained test sample by template comparison and deep learning model, detects bad image and lacks
The specific location that defect occurs in pixel region and image is fallen into, predicts bad image deflects classification.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201811123897.1A CN109472769A (en) | 2018-09-26 | 2018-09-26 | A kind of bad image defect detection method and system |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201811123897.1A CN109472769A (en) | 2018-09-26 | 2018-09-26 | A kind of bad image defect detection method and system |
Publications (1)
Publication Number | Publication Date |
---|---|
CN109472769A true CN109472769A (en) | 2019-03-15 |
Family
ID=65664666
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201811123897.1A Pending CN109472769A (en) | 2018-09-26 | 2018-09-26 | A kind of bad image defect detection method and system |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN109472769A (en) |
Cited By (16)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110047073A (en) * | 2019-05-05 | 2019-07-23 | 北京大学 | A kind of X-ray weld image fault grading method and system |
CN110288586A (en) * | 2019-06-28 | 2019-09-27 | 昆明能讯科技有限责任公司 | A kind of multiple dimensioned transmission line of electricity defect inspection method based on visible images data |
CN110310262A (en) * | 2019-06-19 | 2019-10-08 | 上海理工大学 | A kind of method, apparatus and system for detection wheel tyre defect |
CN110349145A (en) * | 2019-07-09 | 2019-10-18 | 京东方科技集团股份有限公司 | Defect inspection method, device, electronic equipment and storage medium |
CN110362437A (en) * | 2019-07-16 | 2019-10-22 | 张家港钛思科技有限公司 | The automatic method of embedded device defect tracking based on deep learning |
CN110837856A (en) * | 2019-10-31 | 2020-02-25 | 深圳市商汤科技有限公司 | Neural network training and target detection method, device, equipment and storage medium |
CN111080613A (en) * | 2019-12-12 | 2020-04-28 | 哈尔滨市科佳通用机电股份有限公司 | Image recognition method for damage fault of wagon bathtub |
CN111415330A (en) * | 2020-02-27 | 2020-07-14 | 苏州杰锐思智能科技股份有限公司 | Copper foil appearance defect detection method based on deep learning |
CN111681231A (en) * | 2020-06-10 | 2020-09-18 | 厦门理工学院 | Industrial welding image defect detection method based on target detection |
CN111951218A (en) * | 2020-07-09 | 2020-11-17 | 黄鹏 | Lipstick quality inspection system of mixed deep learning model |
CN112016575A (en) * | 2019-05-29 | 2020-12-01 | 阿里巴巴集团控股有限公司 | Detection method, computer equipment and storage medium |
CN112561866A (en) * | 2020-12-04 | 2021-03-26 | 重庆忽米网络科技有限公司 | Semiconductor substrate photoresist layer defect detection system based on AI and cloud computing technology |
CN112884743A (en) * | 2021-02-22 | 2021-06-01 | 深圳中科飞测科技股份有限公司 | Detection method and device, detection equipment and storage medium |
US11164306B2 (en) | 2019-12-09 | 2021-11-02 | International Business Machines Corporation | Visualization of inspection results |
CN113588672A (en) * | 2021-09-29 | 2021-11-02 | 武汉绿色塑料包装有限公司 | Quality detection method for plastic product |
CN113911890A (en) * | 2021-11-02 | 2022-01-11 | 上海爱登堡电梯集团股份有限公司 | Escalator step missing detection method |
Citations (19)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102694956A (en) * | 2011-01-21 | 2012-09-26 | 株式会社理光 | Image processing apparatus and pixel interpolation method |
CN103234976A (en) * | 2013-04-03 | 2013-08-07 | 江南大学 | Warp knitting machine cloth flaw on-line visual inspection method based on Gabor transformation |
CN104408477A (en) * | 2014-12-18 | 2015-03-11 | 成都铁安科技有限责任公司 | Key part fault detection method and device |
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 |
CN106198569A (en) * | 2016-08-03 | 2016-12-07 | 广东工业大学 | A kind of LTPS/IGZO glass substrate broken hole method for quick |
CN106204618A (en) * | 2016-07-20 | 2016-12-07 | 南京文采科技有限责任公司 | Product surface of package defects detection based on machine vision and sorting technique |
CN106504238A (en) * | 2016-10-31 | 2017-03-15 | 成都交大光芒科技股份有限公司 | Railway contact line defect inspection method based on image procossing and convolutional neural networks |
CN106645190A (en) * | 2017-01-20 | 2017-05-10 | 河北工业大学 | Template matching based precise detection method for steel strip surface defects |
CN106952250A (en) * | 2017-02-28 | 2017-07-14 | 北京科技大学 | A kind of metal plate and belt detection method of surface flaw and device based on Faster R CNN networks |
CN106952257A (en) * | 2017-03-21 | 2017-07-14 | 南京大学 | A kind of curved surface label open defect detection method based on template matches and Similarity Measure |
CN107228860A (en) * | 2017-06-28 | 2017-10-03 | 北京因时机器人科技有限公司 | A kind of gear defect detection method based on image rotation cyclophysis |
CN107463946A (en) * | 2017-07-12 | 2017-12-12 | 浙江大学 | A kind of type of merchandize detection method of combination template matches and deep learning |
CN107543828A (en) * | 2017-08-25 | 2018-01-05 | 广东工业大学 | A kind of Surface Flaw Detection method and system |
CN107833220A (en) * | 2017-11-28 | 2018-03-23 | 河海大学常州校区 | Fabric defect detection method based on depth convolutional neural networks and vision significance |
CN107833210A (en) * | 2017-10-30 | 2018-03-23 | 俊杰机械(深圳)有限公司 | A kind of appearance detecting method and system based on neutral net |
CN107966444A (en) * | 2017-10-12 | 2018-04-27 | 常州信息职业技术学院 | Textile flaw detection method based on template |
CN108345911A (en) * | 2018-04-16 | 2018-07-31 | 东北大学 | Surface Defects in Steel Plate detection method based on convolutional neural networks multi-stage characteristics |
CN108389197A (en) * | 2018-02-26 | 2018-08-10 | 上海赛特斯信息科技股份有限公司 | Transmission line of electricity defect inspection method based on deep learning |
CN108564104A (en) * | 2018-01-09 | 2018-09-21 | 北京百度网讯科技有限公司 | Product defects detection method, device, system, server and storage medium |
-
2018
- 2018-09-26 CN CN201811123897.1A patent/CN109472769A/en active Pending
Patent Citations (19)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102694956A (en) * | 2011-01-21 | 2012-09-26 | 株式会社理光 | Image processing apparatus and pixel interpolation method |
CN103234976A (en) * | 2013-04-03 | 2013-08-07 | 江南大学 | Warp knitting machine cloth flaw on-line visual inspection method based on Gabor transformation |
CN104408477A (en) * | 2014-12-18 | 2015-03-11 | 成都铁安科技有限责任公司 | Key part fault detection method and device |
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 |
CN106204618A (en) * | 2016-07-20 | 2016-12-07 | 南京文采科技有限责任公司 | Product surface of package defects detection based on machine vision and sorting technique |
CN106198569A (en) * | 2016-08-03 | 2016-12-07 | 广东工业大学 | A kind of LTPS/IGZO glass substrate broken hole method for quick |
CN106504238A (en) * | 2016-10-31 | 2017-03-15 | 成都交大光芒科技股份有限公司 | Railway contact line defect inspection method based on image procossing and convolutional neural networks |
CN106645190A (en) * | 2017-01-20 | 2017-05-10 | 河北工业大学 | Template matching based precise detection method for steel strip surface defects |
CN106952250A (en) * | 2017-02-28 | 2017-07-14 | 北京科技大学 | A kind of metal plate and belt detection method of surface flaw and device based on Faster R CNN networks |
CN106952257A (en) * | 2017-03-21 | 2017-07-14 | 南京大学 | A kind of curved surface label open defect detection method based on template matches and Similarity Measure |
CN107228860A (en) * | 2017-06-28 | 2017-10-03 | 北京因时机器人科技有限公司 | A kind of gear defect detection method based on image rotation cyclophysis |
CN107463946A (en) * | 2017-07-12 | 2017-12-12 | 浙江大学 | A kind of type of merchandize detection method of combination template matches and deep learning |
CN107543828A (en) * | 2017-08-25 | 2018-01-05 | 广东工业大学 | A kind of Surface Flaw Detection method and system |
CN107966444A (en) * | 2017-10-12 | 2018-04-27 | 常州信息职业技术学院 | Textile flaw detection method based on template |
CN107833210A (en) * | 2017-10-30 | 2018-03-23 | 俊杰机械(深圳)有限公司 | A kind of appearance detecting method and system based on neutral net |
CN107833220A (en) * | 2017-11-28 | 2018-03-23 | 河海大学常州校区 | Fabric defect detection method based on depth convolutional neural networks and vision significance |
CN108564104A (en) * | 2018-01-09 | 2018-09-21 | 北京百度网讯科技有限公司 | Product defects detection method, device, system, server and storage medium |
CN108389197A (en) * | 2018-02-26 | 2018-08-10 | 上海赛特斯信息科技股份有限公司 | Transmission line of electricity defect inspection method based on deep learning |
CN108345911A (en) * | 2018-04-16 | 2018-07-31 | 东北大学 | Surface Defects in Steel Plate detection method based on convolutional neural networks multi-stage characteristics |
Non-Patent Citations (5)
Title |
---|
孔令丹: ""柔性电路板缺陷检测算法与***研究"", 《中国优秀硕士学位论文全文数据库 信息科技辑》 * |
张义广等: "《非制冷红外成像导引头》", 30 September 2009, 西北工业大学出版社 * |
张培: ""印刷制版的缺陷检测技术研究"", 《中国优秀硕士学位论文全文数据库 工程科技Ⅰ辑》 * |
艾泳宏: ""制衣裁片缺陷检测及定位***研究"", 《中国优秀硕士学位论文全文数据库 信息科技辑》 * |
韩滔: ""基于机器视觉的打孔纸带孔形缺陷检测***研究"", 《中国优秀硕士学位论文全文数据库 信息科技辑》 * |
Cited By (21)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110047073A (en) * | 2019-05-05 | 2019-07-23 | 北京大学 | A kind of X-ray weld image fault grading method and system |
CN110047073B (en) * | 2019-05-05 | 2021-07-06 | 北京大学 | X-ray weld image defect grading method and system |
CN112016575A (en) * | 2019-05-29 | 2020-12-01 | 阿里巴巴集团控股有限公司 | Detection method, computer equipment and storage medium |
CN110310262A (en) * | 2019-06-19 | 2019-10-08 | 上海理工大学 | A kind of method, apparatus and system for detection wheel tyre defect |
CN110288586A (en) * | 2019-06-28 | 2019-09-27 | 昆明能讯科技有限责任公司 | A kind of multiple dimensioned transmission line of electricity defect inspection method based on visible images data |
CN110349145A (en) * | 2019-07-09 | 2019-10-18 | 京东方科技集团股份有限公司 | Defect inspection method, device, electronic equipment and storage medium |
CN110349145B (en) * | 2019-07-09 | 2022-08-16 | 京东方科技集团股份有限公司 | Defect detection method, defect detection device, electronic equipment and storage medium |
CN110362437A (en) * | 2019-07-16 | 2019-10-22 | 张家港钛思科技有限公司 | The automatic method of embedded device defect tracking based on deep learning |
CN110837856A (en) * | 2019-10-31 | 2020-02-25 | 深圳市商汤科技有限公司 | Neural network training and target detection method, device, equipment and storage medium |
US11164306B2 (en) | 2019-12-09 | 2021-11-02 | International Business Machines Corporation | Visualization of inspection results |
CN111080613B (en) * | 2019-12-12 | 2020-11-03 | 哈尔滨市科佳通用机电股份有限公司 | Image recognition method for damage fault of wagon bathtub |
CN111080613A (en) * | 2019-12-12 | 2020-04-28 | 哈尔滨市科佳通用机电股份有限公司 | Image recognition method for damage fault of wagon bathtub |
CN111415330A (en) * | 2020-02-27 | 2020-07-14 | 苏州杰锐思智能科技股份有限公司 | Copper foil appearance defect detection method based on deep learning |
CN111681231A (en) * | 2020-06-10 | 2020-09-18 | 厦门理工学院 | Industrial welding image defect detection method based on target detection |
CN111681231B (en) * | 2020-06-10 | 2023-04-18 | 厦门理工学院 | Industrial welding image defect detection method based on target detection |
CN111951218A (en) * | 2020-07-09 | 2020-11-17 | 黄鹏 | Lipstick quality inspection system of mixed deep learning model |
CN112561866A (en) * | 2020-12-04 | 2021-03-26 | 重庆忽米网络科技有限公司 | Semiconductor substrate photoresist layer defect detection system based on AI and cloud computing technology |
CN112884743A (en) * | 2021-02-22 | 2021-06-01 | 深圳中科飞测科技股份有限公司 | Detection method and device, detection equipment and storage medium |
CN112884743B (en) * | 2021-02-22 | 2024-03-05 | 深圳中科飞测科技股份有限公司 | Detection method and device, detection equipment and storage medium |
CN113588672A (en) * | 2021-09-29 | 2021-11-02 | 武汉绿色塑料包装有限公司 | Quality detection method for plastic product |
CN113911890A (en) * | 2021-11-02 | 2022-01-11 | 上海爱登堡电梯集团股份有限公司 | Escalator step missing detection method |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN109472769A (en) | A kind of bad image defect detection method and system | |
CN111402203B (en) | Fabric surface defect detection method based on convolutional neural network | |
CN106971152B (en) | Method for detecting bird nest in power transmission line based on aerial images | |
CN106875373A (en) | Mobile phone screen MURA defect inspection methods based on convolutional neural networks pruning algorithms | |
CN109272500B (en) | Fabric classification method based on adaptive convolutional neural network | |
CN108520114B (en) | Textile fabric defect detection model and training method and application thereof | |
CN108710919A (en) | A kind of crack automation delineation method based on multi-scale feature fusion deep learning | |
CN109597968B (en) | SMT big data-based solder paste printing performance influence factor analysis method | |
CN109636772A (en) | The defect inspection method on the irregular shape intermetallic composite coating surface based on deep learning | |
CN111402226A (en) | Surface defect detection method based on cascade convolution neural network | |
CN111768365B (en) | Solar cell defect detection method based on convolution neural network multi-feature fusion | |
CN103927534A (en) | Sprayed character online visual detection method based on convolutional neural network | |
CN108256482A (en) | A kind of face age estimation method that Distributed learning is carried out based on convolutional neural networks | |
CN110766681B (en) | Bamboo strip surface defect detection method based on triple loss network | |
CN109544522A (en) | A kind of Surface Defects in Steel Plate detection method and system | |
CN106599230A (en) | Method and system for evaluating distributed data mining model | |
CN108320306A (en) | Merge the video target tracking method of TLD and KCF | |
CN114972312A (en) | Improved insulator defect detection method based on YOLOv4-Tiny | |
CN110110780A (en) | A kind of picture classification method based on confrontation neural network and magnanimity noise data | |
CN106650823A (en) | Probability extreme learning machine integration-based foam nickel surface defect classification method | |
CN112102224A (en) | Cloth defect identification method based on deep convolutional neural network | |
CN112070712B (en) | Printing defect detection method based on self-encoder network | |
CN104461772A (en) | Method for recovering missed data | |
CN110020691A (en) | LCD screen defect inspection method based on the training of convolutional neural networks confrontation type | |
CN116402769A (en) | High-precision intelligent detection method for textile flaws considering size targets |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
RJ01 | Rejection of invention patent application after publication | ||
RJ01 | Rejection of invention patent application after publication |
Application publication date: 20190315 |