CN110097543B - Hot-rolled strip steel surface defect detection method based on generation type countermeasure network - Google Patents
Hot-rolled strip steel surface defect detection method based on generation type countermeasure network Download PDFInfo
- Publication number
- CN110097543B CN110097543B CN201910338448.7A CN201910338448A CN110097543B CN 110097543 B CN110097543 B CN 110097543B CN 201910338448 A CN201910338448 A CN 201910338448A CN 110097543 B CN110097543 B CN 110097543B
- Authority
- CN
- China
- Prior art keywords
- generator
- image
- hot
- strip steel
- rolled strip
- 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
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/004—Artificial life, i.e. computing arrangements simulating life
- G06N3/006—Artificial life, i.e. computing arrangements simulating life based on simulated virtual individual or collective life forms, e.g. social simulations or particle swarm optimisation [PSO]
-
- 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
- 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
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- General Health & Medical Sciences (AREA)
- General Engineering & Computer Science (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Artificial Intelligence (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Biomedical Technology (AREA)
- Life Sciences & Earth Sciences (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Health & Medical Sciences (AREA)
- Quality & Reliability (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Image Analysis (AREA)
Abstract
The invention relates to a hot-rolled strip steel surface defect detection method based on a generative confrontation network, which comprises the following specific steps: (1) Extracting surface defect images of hot-rolled strip steel on an industrial site, and performing image preprocessing; (2) And constructing a generator model and a discriminator model of the generative confrontation network GAN. Adding a condition label vector c into the input of a generator for outputting a classified image; introducing pixel loss L in generator training p The quality of the generated image is improved; a discrimination branch and a multi-classification branch are arranged in the discriminator, so that the multi-classification function is realized and the classification precision is improved; (3) Optimizing the constructed generative confrontation network parameters by using a Particle Swarm Optimization (PSO); (4) And combining the generated image and the real image into a hot-rolled strip steel surface defect sample set. The method provided by the invention can solve the problem of insufficient sample data, improve the speed and accuracy of defect image feature extraction, and provide a new effective method for the surface defect detection of the hot-rolled strip steel.
Description
Technical Field
The invention belongs to the technical field of computer vision detection, and particularly relates to a hot-rolled strip steel surface defect detection method based on a generative confrontation network.
Technical Field
The surface defects of the strip steel seriously affect the appearance, fatigue strength, corrosion resistance, wear resistance and other properties of the steel product, affect the subsequent use of the steel product and cause immeasurable industrial loss. Therefore, the defect detection of the strip steel product is a very important step in the industrial production. The method for detecting the surface defects of the strip steel is developed from manual detection to machine detection, so that the speed and the identification precision are improved. The most common method for detecting the defects of the strip steel at present is to extract and process the characteristics of the defects by using different means and then classify the defects by using a classifier. In 2002T Ojala et al introduced a classification method based on Uniform Local Binary Patterns (ULBP) that achieved good performance in terms of spatial resolution, gray scale variation, rotation, etc. In 2013, santan Ghorai et al propose an automatic visual inspection system with Discrete Wavelet Transform (DWT) characteristics and a Support Vector Machine (SVM). The Songcheng university in the northeast of 2013 introduces a Local Binary Pattern (LBP) -based band steel defect identification method capable of effectively resisting noise, the average precision is 98%, and the method can be effectively applied to actual production. In 2016, the King Lei et al adopts Tetrol transformation to decompose the surface image of the hot rolled steel plate to obtain a high-dimensional characteristic vector, and sends the high-dimensional characteristic vector into an SVM to finish classification and identification of the surface defects of the hot rolled steel plate, the classification of the method comprises 8 defects such as transverse cracks, and the defect classification accuracy is 97.38%.
A Generative Adaptive Network (GAN) is a Generative model, and its most basic application is to model distribution of real data and generate sample data, such as generating images and videos. The internal antagonistic training mechanism of GAN has outstanding learning ability, can effectively solve the problem of insufficient samples encountered in the traditional machine learning, improves the speed and accuracy of feature extraction, and provides a new idea for identifying the surface defects of the hot-rolled strip steel.
Disclosure of Invention
The invention aims to provide a hot-rolled strip steel surface defect detection method based on a generative countermeasure network, which can solve the problems of few data samples, low fitting degree and the like and improve the identification precision by applying GAN to hot-rolled strip steel surface defect detection.
The technical scheme adopted by the invention is as follows:
a hot-rolled strip steel surface defect detection method based on a generative countermeasure network comprises the following steps:
the method comprises the following steps: and extracting the surface defect image of the hot-rolled strip steel on the industrial site in a certain steel mill according to the condition of the industrial production site, and preprocessing the image.
Step two: and constructing a generation model and a discrimination model of the generative countermeasure network GAN.
The following improvements are made to the generator model in the generative confrontation network. First, a conditional label vector c is added to the generator input, i.e., the generator input consists of random noise z and a conditional label c. Second, pixel loss L is introduced during generator training p And the quality of the generated image is improved. The network model structure of the improved generator is as follows: the input is a vector formed by connecting a random vector and a label vector, and the vector sequentially passes through a convolutional layer Conv, an active layer Relu, a Residual structure Residual Block and a deconvolution layer Deconv.
The discriminator model in the generative countermeasure network is composed of a discrimination branch and a multi-classification branch. Introduction in original GAN discriminatorAnd a conditional label c is used as a judgment branch, and a multi-classification branch is added to form a GAN (generic area network) discriminator together. The judgment branch is used for judging whether the input image is a real image or an image generated by the generator, and the multi-classification branch has two functions, namely judging the classification of the input image and generating the classification loss L cls . The network model of the discrimination branch circuit is similar to that of the multi-classification branch circuit, and the discrimination branch circuit is composed of a convolution layer Conv, an activation layer LeakyRelu and a deconvolution layer Deconv.
Step three: and optimizing the constructed generative countermeasure network parameters by using a Particle Swarm Optimization (PSO).
The training of the generative countermeasure network adopts an alternate training mode, and the training optimization function of the GAN model is
Because the structures of a generator and an arbiter of the GAN model are complex, a generating type confrontation network optimization method based on Particle Swarm Optimization (PSO) is provided. The specific optimization process is to use the parameter trained in the model as a particle in the PSO algorithm, and the length n of the particle is the number of parameters participating in training in the network. And (4) taking the loss function L as a fitness function of the PSO algorithm to obtain a local optimal solution and a global optimal solution of the particles. And updating the positions and the speeds of the particles in an iteration mode to obtain the particles which are the updated network weight, and iterating until an adaptive value, namely a training error, converges to a minimum value of a threshold range to complete parameter optimization of a generator and a discriminator of the generation type confrontation network.
Step four: and (3) taking the acquired real hot-rolled strip steel surface defect image and the condition label c as the input of a generator G, and mixing the output image and the real image to be used as a hot-rolled strip steel surface defect sample set. And taking part of the sample pictures as a test sample set, and taking the rest of the sample pictures as a training sample set.
Step five: and extracting the trained discriminator D, removing the last layer, carrying out structure fine adjustment, and effectively identifying and classifying the hot-rolled strip steel defect sample image data.
Further, the pixel loss L introduced in the generator training p The definition of the output image of the generator can be increased, and the generated high-pixel image is used for improving the identification accuracy of the surface defects of the hot-rolled strip steel. Wherein
L p =|G(z)-x|
Judging loss L of judging branch of the discriminator in the feedback of the full connection layer adv Whether the input to the discriminator is a real image or a generated image. The output of the full-connection layer of the multi-classification branch is an N-dimensional vector, N is the class number of the hot-rolled strip steel surface defect data set, and the training aims to optimize the classification loss L of feedback cls . During training, the discrimination branch and the multi-classification branch share the weight of the upper layer, and the discrimination loss L of the discrimination branch adv And classification loss L of multi-classification branch feedback cls The coaching generator generates a high quality image.
In summary, the loss function of the GAN model is
L=L adv +L cls +L p
Compared with the prior art, the invention has the advantages that:
the method constructs and trains the generating type countermeasure network GAN, can generate the defect image of the hot-rolled strip steel, enriches the diversity and the randomness of a training sample set, effectively enhances the data sample, solves the problem of the lack of the sample data of the surface defect of the hot-rolled strip steel, and avoids the over-fitting phenomenon.
2, the invention improves the generator and the discriminator of the original GAN, introduces pixel loss in the generator, guides to generate a high-resolution defect image, and uses the high-quality image to replace a low-quality image for defect identification; the discrimination branch and the classification branch are added into the discriminator at the same time, so that the authenticity of the image is judged and the effective classification and identification are carried out, the average classification precision of the surface defects of the hot rolled strip steel can be improved, and a better identification effect is obtained.
3, the parameters of the GAN model are optimized by using a Particle Swarm Optimization (PSO), the PSO has stronger global optimization capability, the problems of gradient reduction and the like in the traditional optimization mode are avoided, the generated confrontation network is accelerated to converge, and the training effect is improved.
Drawings
FIG. 1 is a flow chart of a hot-rolled strip steel surface defect detection method based on a generative countermeasure network of the present invention.
Fig. 2 is a schematic structural diagram of the generative countermeasure network of the present invention.
Fig. 3 is a schematic diagram of a training process of the generative confrontation network of the present invention.
Detailed Description
The invention is further described with reference to the following figures and detailed description.
As shown in fig. 1, the implementation process of the present invention is as follows:
step one, acquiring a defect image of the hot-rolled strip steel on an industrial site, and performing primary image preprocessing.
And step two, constructing and optimizing a generative confrontation network model, inputting the processed real strip steel surface defect image and the condition label into the model, and observing the output generated image.
And step three, mixing the generated image with the collected hot-rolled strip steel surface defect image, and performing image processing to obtain a hot-rolled strip steel surface defect sample set.
And step four, training the improved GAN by using a particle swarm algorithm, extracting a discrimination model as a classifier, adjusting parameters, effectively identifying the image data of the hot-rolled strip steel defect sample and classifying the image data.
Fig. 2 is a schematic structural diagram of the generative countermeasure network of the present invention.
The original GAN is composed of a generator G and a discriminator D. Wherein the generator G generates pseudo data similar to the real data based on the random noise, and the discriminator D judges whether the input is the real data or the generated pseudo data. Because the original GAN has a simple structure, is difficult to control and is trained too freely, a conditional label c is introduced into a generation model and a discrimination model in the GAN model of the research, and defect category information is contained in the label to guide the data generation of the generator and the recognition and classification of a discriminator. In the discriminator model, the discriminator is divided into two branches, namely a discrimination branch and a multi-classification branch, so that the functions of discriminating data true and false and identifying classification are realized respectively.
Fig. 3 is a schematic diagram illustrating a training process of the generative countermeasure network according to the present invention.
And (3) a training optimization process of the original GAN is a maximin and minimum game process of the generator G and the discriminator D. Thus the model optimization formula can be written as
In order to better distinguish whether the data is real data or data generated by the generator G, the discriminator D needs to accurately identify the two and differentiate the probability of output after judgment as much as possible, i.e., D (x) is as large as possible and D (G (x)) is as small as possible; the generator G needs to continuously increase the similarity between the generated data and the real data, so that the generated data and the real data cannot be distinguished by the discriminator. Therefore, the training process of the two modules is a process of mutual competition and mutual confrontation, the performances of the two modules are improved in continuous iteration, and finally Nash balance in the game theory is achieved (namely the discrimination probability of generated data is the same as that of real data), and at the moment, the discriminator D and the generator G are optimal.
In the improved GAN practical optimization process of the present study, a step-by-step alternate training method is adopted for GAN training, and the specific training steps are as follows:
(1) A generator network is trained. And inputting the acquired real defect data x mixed condition label c into a generator to generate a high-quality image. Calculating pixel loss L p The greater the pixel loss, the greater the difference between the high quality image and the real data. The process of training the generator is to minimize the pixel loss L by using the particle swarm algorithm p ;
(2) And fixing the parameters of the generator, and training a distinguishing branch and a classifying branch of the discriminator by using the high-quality image sample generated by the generator.
(3) Fixing the branch discriminating parameters of the discriminator, inputting the real data x into the generator, outputting high-quality image, calculating the loss L of the branch discriminating adv And feeding back to the generator and optimizing parameters of the generator.
(4) Fixing the branch parameters of the classifier, inputting the real data x into the generator, outputting high-quality image, and calculating the loss L of the branch cls Feeding back to the generator and optimizing the parameters of the generator.
Repeating the steps (2) to (4) and performing iterative training 10000 times.
The foregoing description of specific embodiments of the present invention has been presented. It is to be understood that the present invention is not limited to the specific embodiments described above, and that various changes or modifications may be made by one skilled in the art within the scope of the appended claims without departing from the spirit of the invention. The embodiments and features of the embodiments of the present application may be combined with each other arbitrarily without conflict.
Claims (2)
1. A hot-rolled strip steel surface defect detection method based on a generative countermeasure network is characterized by comprising the following steps:
step one, extracting surface defect images of hot-rolled strip steel on an industrial site in a steel mill by combining the condition of the industrial production site, and preprocessing the images;
step two, constructing a generator model and a discriminator model of the generative countermeasure network GAN;
the generator model in the generative confrontation network is improved as follows;
(1) Adding a conditional label vector c into the input of the generator, namely the input of the generator consists of random noise z and a conditional label c;
(2) Introducing pixel loss L during generator training p Improving the quality of the generated image; the network model structure of the improved generator is as follows: the input is a vector formed by connecting a random vector and a label vector, and the vector sequentially passes through a convolution layer Conv, an activation layer Relu, a Residual structure Residual Block and a deconvolution layer Deconv;
the discriminator model in the generative countermeasure network is improved as follows:
introducing a conditional label c as a discrimination branch in a discriminator of the original GAN, and adding oneA plurality of classification branches jointly form a GAN discriminator; the judgment branch is used for judging whether the input image is a real image or an image generated by the generator, and the multi-classification branch has two functions, namely judging the classification of the input image and generating the classification loss L cls (ii) a The network models of the discrimination branch and the multi-classification branch are similar and are composed of a convolution layer Conv, an active layer LeakyRelu and a deconvolution layer Deconv;
step three, optimizing the constructed generative countermeasure network parameters by using a Particle Swarm Optimization (PSO);
the training of the generative countermeasure network adopts an alternate training mode, and the training optimization function of the GAN model is
Because the structures of a generator and a discriminator of the GAN model are complex, a generating type confrontation network optimization method based on particle swarm optimization PSO is provided; the specific optimization process is that the parameter trained in the model is used as a particle in the PSO algorithm, and the length n of the particle is the number of parameters participating in training in the network; taking a loss function L as a fitness function of a PSO algorithm to obtain a local optimal solution and a global optimal solution of the particles; updating the positions and the speeds of the particles in an iteration mode, obtaining the particles which are updated network weights, iterating until an adaptive value, namely a training error converges to a minimum value of a threshold range, and determining according to a data sample to complete parameter optimization of a generator and a discriminator of the generation type confrontation network;
step four, the collected real hot-rolled strip steel surface defect image and the condition label c are used as the input of a generator G, and the output image and the real image are mixed to be used as a hot-rolled strip steel surface defect sample set; taking part of sample pictures as a test sample set, and taking the rest as a training sample set;
and step five, extracting the trained discriminator D, removing the last layer, carrying out structure fine adjustment, and effectively identifying and classifying the hot-rolled strip steel defect sample image data.
2. The method for detecting the surface defects of the hot-rolled strip steel based on the generative countermeasure network as claimed in claim 1, wherein the pixel loss L introduced in the training of the generator p The definition of the output image of the generator can be increased, and the generated high-pixel image is used for improving the identification accuracy of the surface defects of the hot-rolled strip steel; wherein
L p =|G(z)-x|
Judging loss L of judging branch of the discriminator in the feedback of the full connection layer adv For characterizing whether the input of the discriminator is a real image or a generated image; the output of the full connecting layer of the multi-classification branch is an N-dimensional vector, and N is the category number of the surface defect data set of the hot-rolled strip steel; the purpose of its training is to optimize the classification loss L of the feedback cls (ii) a During training, the discrimination branch and the multi-classification branch share the weight of the upper layer, and the discrimination loss L of the discrimination branch adv And classification loss L of multi-classification branch feedback cls The coaching generator generates a high quality image;
in summary, the loss function of the GAN model is:
L=L adv +L cls +L p 。
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910338448.7A CN110097543B (en) | 2019-04-25 | 2019-04-25 | Hot-rolled strip steel surface defect detection method based on generation type countermeasure network |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910338448.7A CN110097543B (en) | 2019-04-25 | 2019-04-25 | Hot-rolled strip steel surface defect detection method based on generation type countermeasure network |
Publications (2)
Publication Number | Publication Date |
---|---|
CN110097543A CN110097543A (en) | 2019-08-06 |
CN110097543B true CN110097543B (en) | 2023-01-13 |
Family
ID=67445823
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201910338448.7A Active CN110097543B (en) | 2019-04-25 | 2019-04-25 | Hot-rolled strip steel surface defect detection method based on generation type countermeasure network |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN110097543B (en) |
Families Citing this family (27)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110706308B (en) * | 2019-09-07 | 2020-09-25 | 创新奇智(成都)科技有限公司 | GAN-based steel coil end face edge loss artificial sample generation method |
CN112488984A (en) * | 2019-09-11 | 2021-03-12 | 中信戴卡股份有限公司 | Method and device for acquiring defect picture generation network and defect picture generation method |
CN110796174A (en) * | 2019-09-29 | 2020-02-14 | 郑州金惠计算机系统工程有限公司 | Multi-type virtual sample generation method and device, electronic equipment and storage medium |
CN112718881A (en) * | 2019-10-28 | 2021-04-30 | 南京智欧智能技术研究院有限公司 | Method and device for detecting surface defects of hot-rolled steel material |
CN111126433A (en) * | 2019-11-19 | 2020-05-08 | 佛山市南海区广工大数控装备协同创新研究院 | Positive and negative sample data balancing method in factory PCB defect detection |
CN111126446B (en) * | 2019-11-29 | 2023-04-07 | 西安工程大学 | Method for amplifying defect image data of robot vision industrial product |
CN111126566B (en) * | 2019-11-29 | 2022-07-08 | 江苏艾佳家居用品有限公司 | Abnormal furniture layout data detection method based on GAN model |
CN112990418B (en) * | 2019-12-02 | 2024-05-07 | 国网能源研究院有限公司 | Multi-energy cross-region optimal configuration method based on particle swarm-linear programming algorithm |
CN111062918B (en) * | 2019-12-10 | 2023-11-21 | 歌尔股份有限公司 | Abnormality detection method and device based on computer vision |
WO2021158214A1 (en) * | 2020-02-05 | 2021-08-12 | Google Llc | Image transformation using interpretable transformation parameters |
CN111223093A (en) * | 2020-03-04 | 2020-06-02 | 武汉精立电子技术有限公司 | AOI defect detection method |
CN111652209B (en) * | 2020-04-30 | 2024-05-10 | 平安科技(深圳)有限公司 | Damage detection method, device, electronic equipment and medium |
CN111986142A (en) * | 2020-05-23 | 2020-11-24 | 冶金自动化研究设计院 | Unsupervised enhancement method for surface defect image data of hot-rolled plate coil |
CN111652864A (en) * | 2020-05-29 | 2020-09-11 | 沈阳铸造研究所有限公司 | Casting defect image generation method for generating countermeasure network based on conditional expression |
US11328410B2 (en) * | 2020-08-03 | 2022-05-10 | KLA Corp. | Deep generative models for optical or other mode selection |
CN112215055A (en) * | 2020-08-15 | 2021-01-12 | 天津大学 | Method for detecting conductor defects in power transmission line based on single classification method |
CN112132181B (en) * | 2020-08-20 | 2023-05-05 | 黑龙江大学 | Image true and false identification method based on generation type countermeasure network |
CN112036626B (en) * | 2020-08-21 | 2023-08-22 | 燕山大学 | Online forecasting method for edge linear defects of hot rolled strip steel |
CN112085734B (en) * | 2020-09-25 | 2022-02-01 | 西安交通大学 | GAN-based image restoration defect detection method |
CN112418310A (en) * | 2020-11-20 | 2021-02-26 | 第四范式(北京)技术有限公司 | Text style migration model training method and system and image generation method and system |
CN113298190B (en) * | 2021-07-05 | 2023-04-07 | 四川大学 | Weld image recognition and classification algorithm based on large-size unbalanced samples |
CN113469913B (en) * | 2021-07-06 | 2022-06-03 | 中南大学 | Hot-rolled strip steel surface water drop removing method based on gradual cycle generation countermeasure network |
CN113450344B (en) * | 2021-07-19 | 2022-05-13 | 江南大学 | Strip steel surface defect detection method and system |
CN113706477B (en) * | 2021-08-10 | 2024-02-13 | 南京旭锐软件科技有限公司 | Defect category identification method, device, equipment and medium |
CN113838208B (en) * | 2021-09-02 | 2024-02-02 | 桂林电子科技大学 | Strip steel defect detection method based on improved residual shrinkage network |
CN114511570A (en) * | 2022-04-20 | 2022-05-17 | 深圳市信润富联数字科技有限公司 | Pseudo defect library establishing method, defect detecting method, device and storage medium |
CN117372438A (en) * | 2023-12-08 | 2024-01-09 | 湘江实验室 | Hot rolled steel plate defect detection method based on class unbalance and related equipment |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2018028255A1 (en) * | 2016-08-11 | 2018-02-15 | 深圳市未来媒体技术研究院 | Image saliency detection method based on adversarial network |
CN108388925A (en) * | 2018-03-06 | 2018-08-10 | 天津工业大学 | The anti-pattern collapse robust image generation method for generating network is fought based on New Conditions |
WO2019015466A1 (en) * | 2017-07-17 | 2019-01-24 | 广州广电运通金融电子股份有限公司 | Method and apparatus for verifying person and certificate |
CN109544555A (en) * | 2018-11-26 | 2019-03-29 | 陕西师范大学 | Fine cracks dividing method based on production confrontation network |
-
2019
- 2019-04-25 CN CN201910338448.7A patent/CN110097543B/en active Active
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2018028255A1 (en) * | 2016-08-11 | 2018-02-15 | 深圳市未来媒体技术研究院 | Image saliency detection method based on adversarial network |
WO2019015466A1 (en) * | 2017-07-17 | 2019-01-24 | 广州广电运通金融电子股份有限公司 | Method and apparatus for verifying person and certificate |
CN108388925A (en) * | 2018-03-06 | 2018-08-10 | 天津工业大学 | The anti-pattern collapse robust image generation method for generating network is fought based on New Conditions |
CN109544555A (en) * | 2018-11-26 | 2019-03-29 | 陕西师范大学 | Fine cracks dividing method based on production confrontation network |
Also Published As
Publication number | Publication date |
---|---|
CN110097543A (en) | 2019-08-06 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN110097543B (en) | Hot-rolled strip steel surface defect detection method based on generation type countermeasure network | |
Ferguson et al. | Automatic localization of casting defects with convolutional neural networks | |
CN110232675B (en) | Texture surface defect detection and segmentation device and method in industrial environment | |
CN108765412B (en) | Strip steel surface defect classification method | |
CN107657279B (en) | Remote sensing target detection method based on small amount of samples | |
Chen et al. | A new ensemble approach based on deep convolutional neural networks for steel surface defect classification | |
CN109766835B (en) | SAR target recognition method for generating countermeasure network based on multi-parameter optimization | |
CN109272500B (en) | Fabric classification method based on adaptive convolutional neural network | |
CN107563999A (en) | A kind of chip defect recognition methods based on convolutional neural networks | |
CN111985499B (en) | High-precision bridge apparent disease identification method based on computer vision | |
CN110197205A (en) | A kind of image-recognizing method of multiple features source residual error network | |
CN111860171A (en) | Method and system for detecting irregular-shaped target in large-scale remote sensing image | |
CN108764298A (en) | Electric power image-context based on single classifier influences recognition methods | |
CN111145145B (en) | Image surface defect detection method based on MobileNet | |
CN112488211A (en) | Fabric image flaw classification method | |
CN113642621A (en) | Zero sample image classification method based on generation countermeasure network | |
CN116883393B (en) | Metal surface defect detection method based on anchor frame-free target detection algorithm | |
CN113505810A (en) | Pooling vision-based method for detecting weed growth cycle by using Transformer | |
Ning et al. | Research on surface defect detection algorithm of strip steel based on improved YOLOV3 | |
CN111275684A (en) | Strip steel surface defect detection method based on multi-scale feature extraction | |
CN114120066A (en) | Small sample steel surface defect classification method based on lightweight network | |
CN107024480A (en) | A kind of stereoscopic image acquisition device | |
CN116468083A (en) | Transformer-based network generation countermeasure method | |
CN113901947A (en) | Intelligent identification method for tire surface flaws under small sample | |
CN111046861B (en) | Method for identifying infrared image, method for constructing identification model and application |
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 |