CN111340791A - Photovoltaic module unsupervised defect detection method based on GAN improved algorithm - Google Patents

Photovoltaic module unsupervised defect detection method based on GAN improved algorithm Download PDF

Info

Publication number
CN111340791A
CN111340791A CN202010135948.3A CN202010135948A CN111340791A CN 111340791 A CN111340791 A CN 111340791A CN 202010135948 A CN202010135948 A CN 202010135948A CN 111340791 A CN111340791 A CN 111340791A
Authority
CN
China
Prior art keywords
image
network
training
module
generator
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
Application number
CN202010135948.3A
Other languages
Chinese (zh)
Inventor
寿春晖
洪凌
丁莞尔
周剑武
赵春晖
周文浩
蒋羽
刘轩驿
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Zhejiang University ZJU
Zhejiang Energy Group Research Institute Co Ltd
Original Assignee
Zhejiang University ZJU
Zhejiang Energy Group Research Institute Co Ltd
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Zhejiang University ZJU, Zhejiang Energy Group Research Institute Co Ltd filed Critical Zhejiang University ZJU
Priority to CN202010135948.3A priority Critical patent/CN111340791A/en
Publication of CN111340791A publication Critical patent/CN111340791A/en
Pending legal-status Critical Current

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
    • G01—MEASURING; TESTING
    • G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
    • G01N21/00—Investigating or analysing materials by the use of optical means, i.e. using sub-millimetre waves, infrared, visible or ultraviolet light
    • G01N21/84—Systems specially adapted for particular applications
    • G01N21/88—Investigating the presence of flaws or contamination
    • G01N21/8851—Scan or image signal processing specially adapted therefor, e.g. for scan signal adjustment, for detecting different kinds of defects, for compensating for structures, markings, edges
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/22—Matching criteria, e.g. proximity measures
    • 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
    • G06N3/08—Learning methods
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
    • G01N21/00—Investigating or analysing materials by the use of optical means, i.e. using sub-millimetre waves, infrared, visible or ultraviolet light
    • G01N21/84—Systems specially adapted for particular applications
    • G01N21/88—Investigating the presence of flaws or contamination
    • G01N21/8851—Scan or image signal processing specially adapted therefor, e.g. for scan signal adjustment, for detecting different kinds of defects, for compensating for structures, markings, edges
    • G01N2021/8887—Scan or image signal processing specially adapted therefor, e.g. for scan signal adjustment, for detecting different kinds of defects, for compensating for structures, markings, edges based on image processing techniques
    • 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
    • G06T2207/30164—Workpiece; Machine component

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Data Mining & Analysis (AREA)
  • General Engineering & Computer Science (AREA)
  • Health & Medical Sciences (AREA)
  • Artificial Intelligence (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Evolutionary Computation (AREA)
  • General Health & Medical Sciences (AREA)
  • Computational Linguistics (AREA)
  • Biomedical Technology (AREA)
  • Molecular Biology (AREA)
  • Mathematical Physics (AREA)
  • Software Systems (AREA)
  • Biophysics (AREA)
  • Computing Systems (AREA)
  • Evolutionary Biology (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Signal Processing (AREA)
  • Chemical & Material Sciences (AREA)
  • Analytical Chemistry (AREA)
  • Biochemistry (AREA)
  • Immunology (AREA)
  • Pathology (AREA)
  • Quality & Reliability (AREA)
  • Image Analysis (AREA)

Abstract

The invention relates to a photovoltaic module unsupervised defect detection method based on a GAN improved algorithm, which comprises the following steps: step 1, training and generating an confrontation network module according to an SSIM-GAN algorithm model; step 2, training a coder network module according to an SSIM-GAN algorithm model; step 3, constructing a defect judging module; and 4, detecting the image. The invention has the beneficial effects that: the method can effectively detect tiny and various defects of the photovoltaic module, and can solve the problem of sample imbalance; an SSIM-GAN model is constructed, and the difference between images is described by using structural similarity; the method comprises the steps of generating a normal image through a countermeasure network, mapping the image to a corresponding hidden space through an encoder network, judging whether the image to be detected has defects through a defect detection module, realizing quick and accurate detection of unknown defects under the condition of few defective images, and having the advantages of strong environmental adaptability and robustness.

Description

Photovoltaic module unsupervised defect detection method based on GAN improved algorithm
Technical Field
The invention belongs to the field of industrial product production detection, and particularly relates to a photovoltaic module unsupervised defect detection method based on a GAN improved algorithm.
Background
Due to the gradual depletion of conventional energy sources, new energy sources typified by solar photovoltaic power generation have been rapidly developed in recent years. In the production and processing process of the photovoltaic cell assembly, besides the defects of the material, the damage rate of the cell can be increased by processing the cell for many times on an automatic production line, so that the assembly has the defects of hidden cracks, fragments, insufficient solder joints, grid breakage and the like, and the defects directly influence the conversion efficiency and the service life of the product. With the continuous improvement of the industrial automation level, the traditional manual detection is not suitable for the current automatic production environment due to the defects of low precision, poor real-time performance and high cost.
Machine vision is a widely applied automated detection technology, and the traditional machine vision detection technology needs manual feature extraction and classifier design, which needs to be participated by professionals in the related field and needs to be adapted to algorithms. However, in the actual industrial production process, the industrial parts are various in types and different in specifications, and the types of defects are greatly different according to different industrial products. The traditional machine vision detection technology has the problems of complex design, poor adaptability, low robustness and the like of feature extraction and classifiers, so that the accuracy requirement and the real-time requirement of industrial production cannot be met in industrial detection.
In recent years, deep learning techniques have been widely used in the image field due to their powerful feature learning capabilities. Because the deep learning model does not need to manually extract the features, the feature extraction and the classifier in the traditional machine vision are fused, and the mapping relation from input to output is directly learned. Therefore, complex and complicated image preprocessing and manual feature extraction operations can be avoided, and meanwhile, the method can also adapt to defects of different scenes and different types, so that the industrial detection accuracy, robustness and instantaneity can be improved.
However, in the photovoltaic module data set, abnormal samples are few, and a serious problem of unbalance of positive and negative samples is caused. In the actual production process, the photovoltaic module is not only diversified in abnormality, but also diversified in abnormal performance, and meanwhile, the abnormal area is small. For the photovoltaic hidden crack defect, the defect performance is not obvious, the existing deep learning algorithm usually needs to distribute balanced samples, and can only detect a large object, and the detection capability for tiny abnormality is poor.
Disclosure of Invention
The invention aims to overcome the defects and provides a photovoltaic module unsupervised defect detection method based on a GAN improved algorithm.
The photovoltaic module unsupervised defect detection method based on the GAN improved algorithm comprises the following steps:
step 1, training according to an SSIM-GAN algorithm model to generate an antagonistic network module: training by using normal samples to generate a confrontation network, and obtaining the distribution of the normal samples; the generation countermeasure network module adopts a residual error network model and a WGAN-GP algorithm;
step 2, training a coder network module according to an SSIM-GAN algorithm model: based on the generation countermeasure network module in the step 1, training an encoder network by using the structural similarity and the feature difference of the normal samples to obtain the mapping from the image to the hidden space; the encoder network module adopts a residual error network model;
step 3, constructing a defect judging module: obtaining an abnormal score value based on the structural similarity by using the generation countermeasure network module in the step 1 and the encoder network module in the step 2, and judging whether the image has defects or not by judging whether the abnormal score value reaches a threshold value or not;
step 4, image detection: and inputting the image to be detected into a defect judging module, and returning a judging result by the defect judging module.
Preferably, the process of training and generating the confrontation network module according to the SSIM-GAN algorithm model in step 1 is as follows:
1) collecting photovoltaic module images under actual working conditions;
2) and constructing a data set, wherein the step is realized by the following sub-steps:
2.1), graying the image of the photovoltaic module, and changing the size of the image into a fixed size to obtain a grayscale image with the size of 64 × 64;
2.2) dividing the gray level image into a normal image set and an abnormal image set, wherein the normal image set is a training set I used for generating a countermeasure network module;
3) and constructing and generating a confrontation network model, wherein the step is realized by the following sub-steps:
3.1) constructing a generator residual error network, wherein the input of the generator is 128-dimensional noise generated randomly, the input of the generator comprises 4 residual error blocks, each residual error block consists of two convolution layers of 3 × convolution kernels, and the activation function is Relu;
3.2) constructing a discriminator residual error network, wherein the input of the discriminator is a normal image in a training set and an image generated by a generator, the input of the discriminator comprises 4 residual error blocks, each residual error block consists of two convolution layers of 3 × 3 convolution kernels, and the activation function is Relu;
3.3), the generator and the discriminator form a generation countermeasure network module;
4) training to generate a confrontation network model, wherein the step is realized by the following sub-steps:
4.1), initializing the weight of the generated countermeasure network model into a smaller value by utilizing a Gaussian function, and setting the value of the hyperparameter in the model; the hyper-parameters comprise iteration times, sample set size used by each round of training, learning rate attenuation values and generator and countermeasure training intervals;
4.2) randomly sampling a training set I for generating the countermeasure network module to obtain a batch of normal image samples during the round of training;
4.3), in the round of training, generating 128-dimensional Gaussian noise in the range of 0 to 1;
4.4) inputting the Gaussian noise into a generator residual error network, and outputting an image corresponding to the noise by the generator residual error network;
4.5) inputting the batch of normal image samples and the image output by the generator residual error network into a discriminator residual error network, and respectively calculating a generator and a discriminator loss function, wherein the generator loss function is as follows:
Figure BDA0002397330920000031
in the above formula, f (x) is the value output by the discriminator; x to pgThe representative x is the output value of the generator G; the discriminator loss function is:
Figure BDA0002397330920000032
in the above formula, x to prRepresenting an image where x is true;
4.6) according to the generator loss function and the discriminator loss function, adopting root mean square propagation as an optimizer, alternately training a generator residual error network and a discriminator residual error network, and updating the weight of the generated countermeasure network;
4.7), judging whether the iteration times are reached: if yes, executing step 4.8); if not, returning to the step 4.2);
4.8), saving parameters in the generated countermeasure network and well-trained weights.
Preferably, the process of training the encoder network module according to the SSIM-GAN algorithm model in step 2 is as follows:
1) reading a generation countermeasure network in the trained SSIM-GAN algorithm model;
2) the method comprises the steps of constructing an encoder network module, wherein the input of an encoder network is a training set I which comprises 4 residual blocks, each residual block consists of two convolution layers of 3 × 3 convolution kernels, an activation function is Relu, and the output is a hidden space z;
3) training an encoder network module, wherein the step is realized by the following substeps:
3.1) initializing the weight of the encoder network module to a smaller value by using a Gaussian function, and setting the value of the hyperparameter; the hyper-parameters comprise iteration times, sample set size used by each round of training, learning rate attenuation values and generator and countermeasure training intervals;
3.2) randomly sampling the training set I to obtain a batch of normal image samples x during the training of the round;
3.3) inputting the batch of normal image samples x into an encoder network, and outputting a hidden space E (x) corresponding to the image by the encoder network;
3.4) inputting the hidden space E (x) generated by the image in the step 3.3) into a generator, and outputting a normal image G (E (x)) corresponding to the hidden space;
3.5), calculating a loss function, wherein the step is realized by the following sub-steps:
3.5.1), calculating the structural similarity of the normal image x and the generator generated image G (E (x)); the structural similarity and structural similarity are found according to the following formulas:
Figure BDA0002397330920000041
wherein muxIs the average value of x, μG(E(x))Is G (E:)x)) of the average value of the average values,
Figure BDA0002397330920000042
is the variance of x and is,
Figure BDA0002397330920000043
is the variance, σ, of G (E (x))xG(E(x))Is the covariance of x and G (E (x)), c1And c2Is a constant used to maintain stability;
3.5.2), inputting the normal image x and the generator generated image G (E (x)) into a discriminator, and calculating the squared error loss of the feature output f (x) and f (G (E (x))) of the second last layer of the discriminator:
Lf＝||f(x)-f(G(E(x)))||2(4)
3.5.3), composing an encoder penalty function L from the structural similarity and variance penalties:
L＝-k1SSIM(x，E(x))+k2Lf(5)
in the above formula, k1And k2Is a given proportionality constant;
3.6) training the encoder network by adopting root mean square propagation as an optimizer according to the encoder loss function L, and updating the weight of the encoder network;
3.7), judging whether the iteration times are reached: if yes, executing step 3.8); if not, returning to the step 3.2);
3.8), saving parameters in the encoder network and the trained weight.
Preferably, the process of constructing the defect judgment module in step 3 is as follows:
1) reading a generation countermeasure network in the trained SSIM-GAN algorithm model;
2) reading a coder network in the trained SSIM-GAN algorithm model;
3) and calculating an abnormality discrimination threshold, wherein the step is realized by the following substeps:
3.1) inputting all images x in the training set I into an encoder network to generate a corresponding hidden space E (x);
3.2) inputting the hidden space E (x) in the step 3.1) into the generator to generate a corresponding normal image G (E (x));
3.3), calculating the structural similarity of the x and the G (E (x));
3.4) taking the minimum structural similarity in the training set I as an abnormal discrimination threshold;
4) and constructing a defect judgment module, wherein the step is realized by the following substeps:
4.1), inputting the image to be detected into an encoder network, and generating a corresponding hidden space;
4.2) inputting the hidden space obtained in the step 4.1) into a generator to generate a normal image closest to the hidden space;
4.3) calculating the structural similarity between the image to be detected and the generated normal image;
4.4) judging whether the structural similarity is smaller than a threshold value, if so, judging that the image to be detected is an abnormal image; otherwise, judging the image to be detected as a normal image;
5) and storing the structural similarity threshold in the defect judging module.
Preferably, the image detection process in step 4 is as follows:
1) reading a defect judging module;
2) inputting an image to be detected;
3) preprocessing an image to be detected, wherein the step is realized by the following substeps:
3.1) cutting the image to be detected, removing the background area and obtaining the main part of the image to be detected;
3.2), carrying out graying processing on the detection image, and scaling the detection image to be 64 × 64 in size;
4) and inputting the preprocessed image into a defect judging module, and returning a judging result by the defect judging module.
The invention has the beneficial effects that the invention provides a photovoltaic module unsupervised defect detection method based on a GAN improved algorithm aiming at the defect detection of the photovoltaic module. The method can effectively detect tiny and various defects of the photovoltaic module, and can solve the problem of sample imbalance. An SSIM-GAN model was constructed, and differences between images were described using structural similarity. The method comprises the steps of generating a normal image through a countermeasure network, mapping the image to a corresponding hidden space through an encoder network, judging whether the image to be detected has defects through a defect detection module, realizing quick and accurate detection of unknown defects under the condition of few defective images, and having the advantages of strong environmental adaptability and robustness.
Drawings
FIG. 1 is a schematic representation of the SSIM-GAN model;
FIG. 2 is a schematic diagram of a generation countermeasure network;
FIG. 3 is a flowchart of a generation countermeasure network training process;
FIG. 4 is a schematic diagram of a generator network architecture;
FIG. 5 is a schematic diagram of a network structure of the arbiter;
FIG. 6 is a flow chart of an encoder training process;
FIG. 7 is a schematic diagram of an encoder network architecture;
FIG. 8 is a schematic diagram of a defect detection module;
FIG. 9 is a schematic diagram showing the experimental result of a normal image of an image to be detected;
FIG. 10 is a diagram illustrating an experimental result of an abnormal image of an image to be detected.
Detailed Description
The present invention will be further described with reference to the following examples. The following examples are set forth merely to aid in the understanding of the invention. It should be noted that, for those skilled in the art, it is possible to make various improvements and modifications to the present invention without departing from the principle of the present invention, and those improvements and modifications also fall within the scope of the claims of the present invention.
As shown in fig. 1, the invention provides a photovoltaic module unsupervised defect detection method based on a GAN improved algorithm, and provides an unsupervised anomaly detection model SSIM-GAN capable of detecting tiny and unknown anomalies aiming at the problems of high difficulty in identifying defects of a photovoltaic module, various defect forms, complex detection environment and extreme imbalance of positive and negative samples. The SSIM-GAN algorithm model comprises a generation countermeasure network module, an encoder network module and a defect judgment module. The generation countermeasure network module adopts a residual error network model and a WGAN-GP algorithm, only uses a normal sample to train and generate a countermeasure network, and is used for generating a normal image to obtain the distribution of the normal sample; the encoder network module adopts a residual error network model, trains an encoder network by using normal samples in combination with structural similarity and characteristic difference, and learns the mapping from an image to a hidden space. As shown in fig. 8, the defect determining module obtains an anomaly score value based on structural similarity and feature difference by using the generation countermeasure network module and the encoder network module, and detects a defect by determining whether the anomaly score value reaches a set threshold. In practical application, the image to be detected is input into a defect judging module, and the defect judging module returns a judging result.
The photovoltaic module unsupervised defect detection method based on the GAN improved algorithm comprises the following steps:
step 1, training according to an SSIM-GAN algorithm model to generate an antagonistic network module: training by using normal samples to generate a confrontation network, and obtaining the distribution of the normal samples; the generation countermeasure network module adopts a residual error network model and a WGAN-GP algorithm;
step 2, as shown in fig. 6, training the encoder network module according to the SSIM-GAN algorithm model: based on the generation countermeasure network module in the step 1, training an encoder network by using the structural similarity and the feature difference of the normal samples to obtain the mapping from the image to the hidden space; the encoder network module adopts a residual error network model;
step 3, constructing a defect judging module: obtaining an abnormal score value based on the structural similarity by using the generation countermeasure network module in the step 1 and the encoder network module in the step 2, and judging whether the image has defects or not by judging whether the abnormal score value reaches a threshold value or not;
step 4, image detection, as shown in fig. 9: and inputting the image to be detected into a defect judging module, and returning a judging result by the defect judging module.
As shown in fig. 3, the process of training and generating the countermeasure network module according to the SSIM-GAN algorithm model in step 1 is as follows:
1) collecting photovoltaic module images under actual working conditions;
2) and constructing a data set, wherein the step is realized by the following sub-steps:
2.1), graying the image of the photovoltaic module, and changing the size of the image into a fixed size to obtain a grayscale image with the size of 64 × 64;
2.2) dividing the gray level image into a normal image set and an abnormal image set, wherein the normal image set is a training set I used for generating a countermeasure network module;
3) and constructing and generating a confrontation network model, wherein the step is realized by the following sub-steps:
3.1) constructing a generator residual error network, wherein the input of the generator is 128-dimensional noise generated randomly as shown in FIG. 4, the input of the generator comprises 4 residual error blocks, each residual error block consists of two convolution layers of 3 × convolution kernels, and the activation function is Relu;
3.2) constructing a discriminator residual error network, wherein the input of the discriminator is a normal image in a training set and an image generated by a generator, the input of the discriminator comprises 4 residual error blocks, each residual error block consists of two convolution layers of 3 × 3 convolution kernels, and the activation function is Relu;
3.3), the generator and the discriminator form a generation countermeasure network module;
4) training to generate a confrontation network model, wherein the step is realized by the following sub-steps:
4.1), initializing the weight of the generated countermeasure network model into a smaller value by utilizing a Gaussian function, and setting the value of the hyperparameter in the model; the hyper-parameters comprise iteration times, sample set size used by each round of training, learning rate attenuation values and generator and countermeasure training intervals;
4.2) randomly sampling a training set I for generating the countermeasure network module to obtain a batch of normal image samples during the round of training;
4.3), in the round of training, generating 128-dimensional Gaussian noise in the range of 0 to 1;
4.4) inputting the Gaussian noise into a generator residual error network, and outputting an image corresponding to the noise by the generator residual error network;
4.5) inputting the batch of normal image samples and the image output by the generator residual error network into a discriminator residual error network, and respectively calculating a generator and a discriminator loss function, wherein the generator loss function is as follows:
Figure BDA0002397330920000071
in the above formula, f (x) is the value output by the discriminator; x to pgThe representative x is the output value of the generator G; the discriminator loss function is:
Figure BDA0002397330920000072
in the above formula, x to prRepresenting an image where x is true;
4.6) according to the generator loss function and the discriminator loss function, adopting root mean square propagation as an optimizer, alternately training a generator residual error network and a discriminator residual error network, and updating the weight of the generated countermeasure network;
4.7), judging whether the iteration times are reached: if yes, executing step 4.8); if not, returning to the step 4.2);
4.8), saving parameters in the generated countermeasure network and well-trained weights.
Step 2, the process of training the encoder network module according to the SSIM-GAN algorithm model is as follows:
1) reading a generation countermeasure network in the trained SSIM-GAN algorithm model;
2) as shown in fig. 7, an encoder network module is constructed, wherein the input of the encoder network is a training set I which comprises 4 residual blocks, each residual block is composed of convolution layers of two convolution kernels of 3 × 3, the activation function is Relu, and the output is a hidden space z;
3) training an encoder network module, wherein the step is realized by the following substeps:
3.1) initializing the weight of the encoder network module to a smaller value by using a Gaussian function, and setting the value of the hyperparameter; the hyper-parameters comprise iteration times, sample set size used by each round of training, learning rate attenuation values and generator and countermeasure training intervals;
3.2) randomly sampling the training set I to obtain a batch of normal image samples x during the training of the round;
3.3) inputting the batch of normal image samples x into an encoder network, and outputting a hidden space E (x) corresponding to the image by the encoder network;
3.4) inputting the hidden space E (x) generated by the image in the step 3.3) into a generator, and outputting a normal image G (E (x)) corresponding to the hidden space;
3.5), calculating a loss function, wherein the step is realized by the following sub-steps:
3.5.1), calculating the structural similarity of the normal image x and the generator generated image G (E (x)); the structural similarity and structural similarity are found according to the following formulas:
Figure BDA0002397330920000081
wherein muxIs the average value of x, μG(E(x))Is the average value of G (E (x)),
Figure BDA0002397330920000082
is the variance of x and is,
Figure BDA0002397330920000083
is the variance, σ, of G (E (x))xG(E(x))Is the covariance of x and G (E (x)), c1And c2Is a constant used to maintain stability;
3.5.2), inputting the normal image x and the generator generated image G (E (x)) into a discriminator, and calculating the squared error loss of the feature output f (x) and f (G (E (x))) of the second last layer of the discriminator:
Lf＝||f(x)-f(G(E(x)))||2(4)
3.5.3), composing an encoder penalty function L from the structural similarity and variance penalties:
L＝-k1SSIM(x，E(x))+k2Lf(5)
in the above formula, k1And k2Is a given proportionality constant;
3.6) training the encoder network by adopting root mean square propagation as an optimizer according to the encoder loss function L, and updating the weight of the encoder network;
3.7), judging whether the iteration times are reached: if yes, executing step 3.8); if not, returning to the step 3.2);
3.8), saving parameters in the encoder network and the trained weight.
The process of constructing the defect judging module in the step 3 is as follows:
1) reading a generation countermeasure network in the trained SSIM-GAN algorithm model;
2) reading a coder network in the trained SSIM-GAN algorithm model;
3) and calculating an abnormality discrimination threshold, wherein the step is realized by the following substeps:
3.1) inputting all images x in the training set I into an encoder network to generate a corresponding hidden space E (x);
3.2) inputting the hidden space E (x) in the step 3.1) into the generator to generate a corresponding normal image G (E (x));
3.3), calculating the structural similarity of the x and the G (E (x));
3.4) taking the minimum structural similarity in the training set I as an abnormal discrimination threshold;
4) and constructing a defect judgment module, wherein the step is realized by the following substeps:
4.1), inputting the image to be detected into an encoder network, and generating a corresponding hidden space;
4.2) inputting the hidden space obtained in the step 4.1) into a generator to generate a normal image closest to the hidden space;
4.3) calculating the structural similarity between the image to be detected and the generated normal image;
4.4) judging whether the structural similarity is smaller than a threshold value, if so, judging that the image to be detected is an abnormal image; otherwise, judging the image to be detected as a normal image;
5) and storing the structural similarity threshold in the defect judging module.
The image detection process in step 4 is as follows:
1) reading a defect judging module;
2) inputting an image to be detected;
3) preprocessing an image to be detected, wherein the step is realized by the following substeps:
3.1) cutting the image to be detected, removing the background area and obtaining the main part of the image to be detected;
3.2), carrying out graying processing on the detection image, and scaling the detection image to be 64 × 64 in size;
4) and inputting the preprocessed image into a defect judging module, and returning a judging result by the defect judging module.
The detection result of the invention is shown in fig. 9 and fig. 10, the first line is the image to be detected, the second line is the normal image which is generated by SSIM-GAN and is closest to the first line, and the third line is the abnormal degree of the image to be detected based on the structural similarity threshold value of 0.6. From the results, it can be seen that the structural similarity between the normal image and the abnormal image is greatly different, and the abnormality can be effectively identified.

Claims (5)

1. A photovoltaic module unsupervised defect detection method based on a GAN improved algorithm is characterized by comprising the following steps:
step 1, training according to an SSIM-GAN algorithm model to generate an antagonistic network module: training by using normal samples to generate a confrontation network, and obtaining the distribution of the normal samples; the generation countermeasure network module adopts a residual error network model and a WGAN-GP algorithm;
step 2, training a coder network module according to an SSIM-GAN algorithm model: based on the generation countermeasure network module in the step 1, training an encoder network by using the structural similarity and the feature difference of the normal samples to obtain the mapping from the image to the hidden space; the encoder network module adopts a residual error network model;
step 3, constructing a defect judging module: obtaining an abnormal score value based on the structural similarity by using the generation countermeasure network module in the step 1 and the encoder network module in the step 2, and judging whether the image has defects or not by judging whether the abnormal score value reaches a threshold value or not;
step 4, image detection: and inputting the image to be detected into a defect judging module, and returning a judging result by the defect judging module.
2. The method for unsupervised defect detection of a photovoltaic module based on a GAN improved algorithm as claimed in claim 1, wherein the process of training and generating the countermeasure network module according to SSIM-GAN algorithm model in step 1 is as follows:
1) collecting photovoltaic module images under actual working conditions;
2) and constructing a data set, wherein the step is realized by the following sub-steps:
2.1), graying the image of the photovoltaic module, and changing the size of the image into a fixed size to obtain a grayscale image with the size of 64 × 64;
2.2) dividing the gray level image into a normal image set and an abnormal image set, wherein the normal image set is a training set I used for generating a countermeasure network module;
3) and constructing and generating a confrontation network model, wherein the step is realized by the following sub-steps:
3.1) constructing a generator residual error network, wherein the input of the generator is 128-dimensional noise generated randomly, the input of the generator comprises 4 residual error blocks, each residual error block consists of two convolution layers of 3 × convolution kernels, and the activation function is Relu;
3.2) constructing a discriminator residual error network, wherein the input of the discriminator is a normal image in a training set and an image generated by a generator, the input of the discriminator comprises 4 residual error blocks, each residual error block consists of two convolution layers of 3 × 3 convolution kernels, and the activation function is Relu;
3.3), the generator and the discriminator form a generation countermeasure network module;
4) training to generate a confrontation network model, wherein the step is realized by the following sub-steps:
4.1), initializing the weight of the generated countermeasure network model into a smaller value by utilizing a Gaussian function, and setting the value of the hyperparameter in the model; the hyper-parameters comprise iteration times, sample set size used by each round of training, learning rate attenuation values and generator and countermeasure training intervals;
4.2) randomly sampling a training set I for generating the countermeasure network module to obtain a batch of normal image samples during the round of training;
4.3), in the round of training, generating 128-dimensional Gaussian noise in the range of 0 to 1;
4.4) inputting the Gaussian noise into a generator residual error network, and outputting an image corresponding to the noise by the generator residual error network;
4.5) inputting the batch of normal image samples and the image output by the generator residual error network into a discriminator residual error network, and respectively calculating a generator and a discriminator loss function, wherein the generator loss function is as follows:
Figure FDA0002397330910000021
in the above formula, f (x) is the value output by the discriminator; x to pgThe representative x is the output value of the generator G; the discriminator loss function is:
Figure FDA0002397330910000022
in the above formula, x to prRepresenting an image where x is true;
4.6) according to the generator loss function and the discriminator loss function, adopting root mean square propagation as an optimizer, alternately training a generator residual error network and a discriminator residual error network, and updating the weight of the generated countermeasure network;
4.7), judging whether the iteration times are reached: if yes, executing step 4.8); if not, returning to the step 4.2);
4.8), saving parameters in the generated countermeasure network and well-trained weights.
3. The method for unsupervised defect detection of a photovoltaic module based on a GAN improved algorithm as claimed in claim 2, wherein the step 2 of training the encoder network module according to SSIM-GAN algorithm model comprises the following steps:
1) reading a generation countermeasure network in the trained SSIM-GAN algorithm model;
2) the method comprises the steps of constructing an encoder network module, wherein the input of an encoder network is a training set I which comprises 4 residual blocks, each residual block consists of two convolution layers of 3 × 3 convolution kernels, an activation function is Relu, and the output is a hidden space z;
3) training an encoder network module, wherein the step is realized by the following substeps:
3.1) initializing the weight of the encoder network module to a smaller value by using a Gaussian function, and setting the value of the hyperparameter; the hyper-parameters comprise iteration times, sample set size used by each round of training, learning rate attenuation values and generator and countermeasure training intervals;
3.2) randomly sampling the training set I to obtain a batch of normal image samples x during the training of the round;
3.3) inputting the batch of normal image samples x into an encoder network, and outputting a hidden space E (x) corresponding to the image by the encoder network;
3.4) inputting the hidden space E (x) generated by the image in the step 3.3) into a generator, and outputting a normal image G (E (x)) corresponding to the hidden space;
3.5), calculating a loss function, wherein the step is realized by the following sub-steps:
3.5.1), calculating the structural similarity of the normal image x and the generator generated image G (E (x)); the structural similarity and structural similarity are found according to the following formulas:
Figure FDA0002397330910000031
wherein muxIs the average value of x, μG(E(x))Is the average value of G (E (x)),
Figure FDA0002397330910000032
is the variance of x and is,
Figure FDA0002397330910000033
is the variance, σ, of G (E (x))xG(E(x))Is the covariance of x and G (E (x)), c1And c2Is a constant used to maintain stability;
3.5.2), inputting the normal image x and the generator generated image G (E (x)) into a discriminator, and calculating the squared error loss of the feature output f (x) and f (G (E (x))) of the second last layer of the discriminator:
Lf＝||f(x)-f(G(E(x)))||2(4)
3.5.3), composing an encoder penalty function L from the structural similarity and variance penalties:
L＝-k1SSIM(x，E(x))+k2Lf(5)
in the above formula, k1And k2Is a given proportionality constant;
3.6) training the encoder network by adopting root mean square propagation as an optimizer according to the encoder loss function L, and updating the weight of the encoder network;
3.7), judging whether the iteration times are reached: if yes, executing step 3.8); if not, returning to the step 3.2);
3.8), saving parameters in the encoder network and the trained weight.
4. The method for unsupervised defect detection of a photovoltaic module based on a GAN improved algorithm as claimed in claim 3, wherein the step 3 of constructing the defect discriminating module comprises the following steps:
1) reading a generation countermeasure network in the trained SSIM-GAN algorithm model;
2) reading a coder network in the trained SSIM-GAN algorithm model;
3) and calculating an abnormality discrimination threshold, wherein the step is realized by the following substeps:
3.1) inputting all images x in the training set I into an encoder network to generate a corresponding hidden space E (x);
3.2) inputting the hidden space E (x) in the step 3.1) into the generator to generate a corresponding normal image G (E (x));
3.3), calculating the structural similarity of the x and the G (E (x));
3.4) taking the minimum structural similarity in the training set I as an abnormal discrimination threshold;
4) and constructing a defect judgment module, wherein the step is realized by the following substeps:
4.1), inputting the image to be detected into an encoder network, and generating a corresponding hidden space;
4.2) inputting the hidden space obtained in the step 4.1) into a generator to generate a normal image closest to the hidden space;
4.3) calculating the structural similarity between the image to be detected and the generated normal image;
4.4) judging whether the structural similarity is smaller than a threshold value, if so, judging that the image to be detected is an abnormal image; otherwise, judging the image to be detected as a normal image;
5) and storing the structural similarity threshold in the defect judging module.
5. The photovoltaic module unsupervised defect detection method based on the GAN improved algorithm as claimed in claim 4, wherein the image detection process of step 4 is as follows:
1) reading a defect judging module;
2) inputting an image to be detected;
3) preprocessing an image to be detected, wherein the step is realized by the following substeps:
3.1) cutting the image to be detected, removing the background area and obtaining the main part of the image to be detected;
3.2), carrying out graying processing on the detection image, and scaling the detection image to be 64 × 64 in size;
4) and inputting the preprocessed image into a defect judging module, and returning a judging result by the defect judging module.
CN202010135948.3A 2020-03-02 2020-03-02 Photovoltaic module unsupervised defect detection method based on GAN improved algorithm Pending CN111340791A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202010135948.3A CN111340791A (en) 2020-03-02 2020-03-02 Photovoltaic module unsupervised defect detection method based on GAN improved algorithm

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202010135948.3A CN111340791A (en) 2020-03-02 2020-03-02 Photovoltaic module unsupervised defect detection method based on GAN improved algorithm

Publications (1)

Publication Number Publication Date
CN111340791A true CN111340791A (en) 2020-06-26

Family

ID=71184675

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202010135948.3A Pending CN111340791A (en) 2020-03-02 2020-03-02 Photovoltaic module unsupervised defect detection method based on GAN improved algorithm

Country Status (1)

Country Link
CN (1) CN111340791A (en)

Cited By (34)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN111768398A (en) * 2020-07-06 2020-10-13 北京瑞莱智慧科技有限公司 Defect detection method and device for photovoltaic module, storage medium and computing equipment
CN111986198A (en) * 2020-09-09 2020-11-24 河南信息安全研究院有限公司 Mold residue detection method and device
CN112001903A (en) * 2020-08-21 2020-11-27 深圳市华汉伟业科技有限公司 Defect detection network construction method, abnormality detection method and system, and storage medium
CN112150432A (en) * 2020-09-22 2020-12-29 电子科技大学 Optical excitation infrared nondestructive testing method based on generation countermeasure network
CN112183342A (en) * 2020-09-28 2021-01-05 国网安徽省电力有限公司检修分公司 Comprehensive convertor station defect identification method with template
CN112215803A (en) * 2020-09-15 2021-01-12 昆明理工大学 Aluminum plate eddy current inspection image defect segmentation method based on improved generation countermeasure network
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
CN112446869A (en) * 2020-11-27 2021-03-05 鹏城实验室 Unsupervised industrial product defect detection method and device based on deep learning
CN112488130A (en) * 2020-12-17 2021-03-12 苏州聚悦信息科技有限公司 AI micro-pore wall detection algorithm
CN112529109A (en) * 2020-12-29 2021-03-19 四川长虹电器股份有限公司 Unsupervised multi-model-based anomaly detection method and system
CN112561197A (en) * 2020-12-23 2021-03-26 国网江苏省电力有限公司南京供电分公司 Power data prefetching and caching method with active defense influence range
CN112668623A (en) * 2020-12-22 2021-04-16 中国铁道科学研究院集团有限公司 Method and device for generating binaural pin defect sample based on generation countermeasure network
CN112837295A (en) * 2021-02-01 2021-05-25 北京工业大学 Rubber glove defect detection method based on generation of countermeasure network
CN112950561A (en) * 2021-02-22 2021-06-11 中国地质大学（武汉） Optical fiber end face defect detection method, device and storage medium
CN113298895A (en) * 2021-06-18 2021-08-24 上海交通大学 Convergence guarantee-oriented unsupervised bidirectional generation automatic coding method and system
CN113378703A (en) * 2021-06-09 2021-09-10 成都恒创新星科技有限公司 Method and system for detecting state of garbage can
CN113407425A (en) * 2021-05-13 2021-09-17 桂林电子科技大学 Internal user behavior detection method based on BiGAN and OTSU
CN113449679A (en) * 2021-07-14 2021-09-28 湖南长城科技信息有限公司 Method and device for identifying abnormal behaviors of human body
CN113484333A (en) * 2021-09-08 2021-10-08 苏州高视半导体技术有限公司 Method for distinguishing foreign matter defect of multi-layer structure screen, electronic device and storage medium
CN113570603A (en) * 2021-09-26 2021-10-29 国网浙江省电力有限公司电力科学研究院 Contact network part defect data generation method and system based on cycleGAN
CN113592866A (en) * 2021-09-29 2021-11-02 西安邮电大学 Semiconductor lead frame exposure defect detection method
CN113793343A (en) * 2021-08-23 2021-12-14 南昌航空大学 Defect positioning and segmentation method, system, terminal and storage medium based on image
CN113838040A (en) * 2021-09-29 2021-12-24 西安工程大学 Detection method for defect area of color texture fabric
CN113850335A (en) * 2021-09-28 2021-12-28 哈尔滨工业大学 Data augmentation method for bathroom ceramic defect detection
CN113920096A (en) * 2021-10-14 2022-01-11 广东工业大学 Method for detecting metal packaging defects of integrated circuit
CN114022586A (en) * 2021-10-25 2022-02-08 华中科技大学 Defect image generation method based on countermeasure generation network
CN114022442A (en) * 2021-11-03 2022-02-08 武汉智目智能技术合伙企业(有限合伙) Unsupervised learning-based fabric defect detection algorithm
CN114065874A (en) * 2021-11-30 2022-02-18 河北省科学院应用数学研究所 Medicine glass bottle appearance defect detection model training method and device and terminal equipment
WO2022067876A1 (en) * 2020-09-29 2022-04-07 Hong Kong Applied Science And Technology Research Institute Co. Ltd. Method and system of image based anomaly localization for vehicles through generative contextualized adversarial network
CN114723660A (en) * 2022-03-01 2022-07-08 华中科技大学 Cloth defect detection method and system
CN114820541A (en) * 2022-05-07 2022-07-29 武汉象点科技有限公司 Defect detection method based on reconstructed network
CN114862772A (en) * 2022-04-21 2022-08-05 南京大学 Unsupervised defect detection method based on deep learning feature correlation
CN117197095A (en) * 2023-09-14 2023-12-08 四川轻化工大学 Surface defect detection method and system for generating countermeasure segmentation model based on semi-supervision
CN117315354A (en) * 2023-09-27 2023-12-29 南京航空航天大学 Insulator anomaly detection method based on multi-discriminant composite coding GAN network

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109544555A (en) * 2018-11-26 2019-03-29 陕西师范大学 Fine cracks dividing method based on production confrontation network
CN110211114A (en) * 2019-06-03 2019-09-06 浙江大学 A kind of scarce visible detection method of the vanning based on deep learning
CN110544253A (en) * 2019-09-12 2019-12-06 福州大学 fabric flaw detection method based on image pyramid and column template
CN110555474A (en) * 2019-08-28 2019-12-10 上海电力大学 photovoltaic panel fault detection method based on semi-supervised learning
CN110619146A (en) * 2019-08-12 2019-12-27 河北工业大学 Polycrystalline silicon cell crack defect detection method based on structural similarity measurement

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109544555A (en) * 2018-11-26 2019-03-29 陕西师范大学 Fine cracks dividing method based on production confrontation network
CN110211114A (en) * 2019-06-03 2019-09-06 浙江大学 A kind of scarce visible detection method of the vanning based on deep learning
CN110619146A (en) * 2019-08-12 2019-12-27 河北工业大学 Polycrystalline silicon cell crack defect detection method based on structural similarity measurement
CN110555474A (en) * 2019-08-28 2019-12-10 上海电力大学 photovoltaic panel fault detection method based on semi-supervised learning
CN110544253A (en) * 2019-09-12 2019-12-06 福州大学 fabric flaw detection method based on image pyramid and column template

Non-Patent Citations (3)

* Cited by examiner, † Cited by third party
Title
HAIYONG CHEN 等: "Accurate and robust crack detection using steerable evidence filtering in electroluminescence images of solar cells", 《OPTICS AND LASERS IN ENGINEERING》 *
THOMAS SCHLEGL 等: "f-AnoGAN: Fast unsupervised anomaly detection with generative adversarial networks", 《MEDICAL IMAGE ANALYSISMEDICAL IMAGE ANALYSIS》 *
董伟: "生成对抗网络研究及其在图像风格转换的应用", 《中国优秀博硕士学位论文全文数据库(硕士)信息科技辑》 *

Cited By (48)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN111768398A (en) * 2020-07-06 2020-10-13 北京瑞莱智慧科技有限公司 Defect detection method and device for photovoltaic module, storage medium and computing equipment
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
CN112001903A (en) * 2020-08-21 2020-11-27 深圳市华汉伟业科技有限公司 Defect detection network construction method, abnormality detection method and system, and storage medium
CN111986198A (en) * 2020-09-09 2020-11-24 河南信息安全研究院有限公司 Mold residue detection method and device
CN112215803A (en) * 2020-09-15 2021-01-12 昆明理工大学 Aluminum plate eddy current inspection image defect segmentation method based on improved generation countermeasure network
CN112150432A (en) * 2020-09-22 2020-12-29 电子科技大学 Optical excitation infrared nondestructive testing method based on generation countermeasure network
CN112183342A (en) * 2020-09-28 2021-01-05 国网安徽省电力有限公司检修分公司 Comprehensive convertor station defect identification method with template
CN112183342B (en) * 2020-09-28 2022-07-12 国网安徽省电力有限公司检修分公司 Comprehensive convertor station defect identification method with template
WO2022067876A1 (en) * 2020-09-29 2022-04-07 Hong Kong Applied Science And Technology Research Institute Co. Ltd. Method and system of image based anomaly localization for vehicles through generative contextualized adversarial network
US11328402B2 (en) 2020-09-29 2022-05-10 Hong Kong Applied Science and Technology Research Institute Company Limited Method and system of image based anomaly localization for vehicles through generative contextualized adversarial network
CN112446869A (en) * 2020-11-27 2021-03-05 鹏城实验室 Unsupervised industrial product defect detection method and device based on deep learning
CN112488130B (en) * 2020-12-17 2023-08-15 苏州聚悦信息科技有限公司 AI micro hole wall detection method
CN112488130A (en) * 2020-12-17 2021-03-12 苏州聚悦信息科技有限公司 AI micro-pore wall detection algorithm
CN112668623B (en) * 2020-12-22 2024-04-16 中国铁道科学研究院集团有限公司 Method and device for generating binaural pin defect sample based on generation countermeasure network
CN112668623A (en) * 2020-12-22 2021-04-16 中国铁道科学研究院集团有限公司 Method and device for generating binaural pin defect sample based on generation countermeasure network
CN112561197A (en) * 2020-12-23 2021-03-26 国网江苏省电力有限公司南京供电分公司 Power data prefetching and caching method with active defense influence range
CN112529109A (en) * 2020-12-29 2021-03-19 四川长虹电器股份有限公司 Unsupervised multi-model-based anomaly detection method and system
CN112837295A (en) * 2021-02-01 2021-05-25 北京工业大学 Rubber glove defect detection method based on generation of countermeasure network
CN112950561A (en) * 2021-02-22 2021-06-11 中国地质大学（武汉） Optical fiber end face defect detection method, device and storage medium
CN112950561B (en) * 2021-02-22 2022-07-26 中国地质大学（武汉） Optical fiber end face defect detection method, device and storage medium
CN113407425B (en) * 2021-05-13 2022-09-23 桂林电子科技大学 Internal user behavior detection method based on BiGAN and OTSU
CN113407425A (en) * 2021-05-13 2021-09-17 桂林电子科技大学 Internal user behavior detection method based on BiGAN and OTSU
CN113378703A (en) * 2021-06-09 2021-09-10 成都恒创新星科技有限公司 Method and system for detecting state of garbage can
CN113298895A (en) * 2021-06-18 2021-08-24 上海交通大学 Convergence guarantee-oriented unsupervised bidirectional generation automatic coding method and system
CN113298895B (en) * 2021-06-18 2023-05-12 上海交通大学 Automatic encoding method and system for unsupervised bidirectional generation oriented to convergence guarantee
CN113449679A (en) * 2021-07-14 2021-09-28 湖南长城科技信息有限公司 Method and device for identifying abnormal behaviors of human body
CN113793343B (en) * 2021-08-23 2023-06-06 南昌航空大学 Image-based defect positioning and segmentation method, system, terminal and storage medium
CN113793343A (en) * 2021-08-23 2021-12-14 南昌航空大学 Defect positioning and segmentation method, system, terminal and storage medium based on image
CN113484333A (en) * 2021-09-08 2021-10-08 苏州高视半导体技术有限公司 Method for distinguishing foreign matter defect of multi-layer structure screen, electronic device and storage medium
CN113484333B (en) * 2021-09-08 2021-12-14 苏州高视半导体技术有限公司 Method for distinguishing foreign matter defect of multi-layer structure screen, electronic device and storage medium
CN113570603A (en) * 2021-09-26 2021-10-29 国网浙江省电力有限公司电力科学研究院 Contact network part defect data generation method and system based on cycleGAN
CN113850335A (en) * 2021-09-28 2021-12-28 哈尔滨工业大学 Data augmentation method for bathroom ceramic defect detection
CN113592866A (en) * 2021-09-29 2021-11-02 西安邮电大学 Semiconductor lead frame exposure defect detection method
CN113838040A (en) * 2021-09-29 2021-12-24 西安工程大学 Detection method for defect area of color texture fabric
CN113920096A (en) * 2021-10-14 2022-01-11 广东工业大学 Method for detecting metal packaging defects of integrated circuit
CN114022586A (en) * 2021-10-25 2022-02-08 华中科技大学 Defect image generation method based on countermeasure generation network
CN114022586B (en) * 2021-10-25 2024-07-02 华中科技大学 Defect image generation method based on countermeasure generation network
CN114022442A (en) * 2021-11-03 2022-02-08 武汉智目智能技术合伙企业(有限合伙) Unsupervised learning-based fabric defect detection algorithm
CN114065874B (en) * 2021-11-30 2024-08-16 河北省科学院应用数学研究所 Training method and device for appearance defect detection model of medical glass bottle and terminal equipment
CN114065874A (en) * 2021-11-30 2022-02-18 河北省科学院应用数学研究所 Medicine glass bottle appearance defect detection model training method and device and terminal equipment
CN114723660A (en) * 2022-03-01 2022-07-08 华中科技大学 Cloth defect detection method and system
CN114723660B (en) * 2022-03-01 2024-09-10 华中科技大学 Cloth defect detection method and system
CN114862772A (en) * 2022-04-21 2022-08-05 南京大学 Unsupervised defect detection method based on deep learning feature correlation
CN114820541A (en) * 2022-05-07 2022-07-29 武汉象点科技有限公司 Defect detection method based on reconstructed network
CN117197095A (en) * 2023-09-14 2023-12-08 四川轻化工大学 Surface defect detection method and system for generating countermeasure segmentation model based on semi-supervision
CN117197095B (en) * 2023-09-14 2024-04-16 四川轻化工大学 Surface defect detection method and system for generating countermeasure segmentation model based on semi-supervision
CN117315354A (en) * 2023-09-27 2023-12-29 南京航空航天大学 Insulator anomaly detection method based on multi-discriminant composite coding GAN network
CN117315354B (en) * 2023-09-27 2024-04-02 南京航空航天大学 Insulator anomaly detection method based on multi-discriminant composite coding GAN network

Similar Documents

Publication Publication Date Title
CN111340791A (en) Photovoltaic module unsupervised defect detection method based on GAN improved algorithm
CN101996405B (en) Method and device for rapidly detecting and classifying defects of glass image
CN111652857B (en) Infrared detection method for insulator defects
CN114619292B (en) Milling cutter wear monitoring method based on fusion of wavelet denoising and attention mechanism with GRU network
CN115861190A (en) Comparison learning-based unsupervised defect detection method for photovoltaic module
CN111652853A (en) Magnetic powder flaw detection method based on deep convolutional neural network
CN113469951B (en) Hub defect detection method based on cascade region convolutional neural network
CN110119747A (en) A kind of coal rock detection method based on radioscopic image
CN108510001B (en) Wind driven generator blade defect classification method and classification system thereof
CN111079978B (en) Coal and gas outburst prediction method based on logistic regression and reinforcement learning
CN113120555B (en) Conveying belt longitudinal tearing detection method based on line structured light
CN104751474A (en) Cascade quick image defect segmentation method
CN109120230B (en) A kind of solar battery sheet EL image detection and defect identification method
CN107316287A (en) A kind of defect identification method in rectangle ferrite magnetic disk sheet face
CN116740044B (en) Copper pipe milling surface processing method and system based on visual detection and control
CN115294033A (en) Tire belt layer difference level and misalignment defect detection method based on semantic segmentation network
CN117269168B (en) New energy automobile precision part surface defect detection device and detection method
CN111179239A (en) Tire X-ray flaw detection method for performing re-ranking by using background features
CN106406257A (en) Iron ore flotation concentrate grade soft measurement method and system based on case-based reasoning
CN112381794B (en) Printing defect detection method based on deep convolution generation network
CN111695582A (en) Flutter texture detection method and device
CN116993711A (en) Self-distillation-based intelligent detection method for cutter damage
CN115311238A (en) Photovoltaic module defect damage identification method based on image analysis
CN113657310A (en) Photovoltaic module infrared hot spot fault identification method based on multilayer convolutional neural network
CN108805855B (en) Method for quickly identifying concrete cracks

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
WD01 Invention patent application deemed withdrawn after publication
WD01 Invention patent application deemed withdrawn after publication

Application publication date: 20200626