CN111291814B - Crack identification algorithm based on convolutional neural network and information entropy data fusion strategy - Google Patents
Crack identification algorithm based on convolutional neural network and information entropy data fusion strategy Download PDFInfo
- Publication number
- CN111291814B CN111291814B CN202010094093.4A CN202010094093A CN111291814B CN 111291814 B CN111291814 B CN 111291814B CN 202010094093 A CN202010094093 A CN 202010094093A CN 111291814 B CN111291814 B CN 111291814B
- Authority
- CN
- China
- Prior art keywords
- crack
- image
- neural network
- convolutional neural
- information entropy
- 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
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- 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
- G01N21/62—Systems in which the material investigated is excited whereby it emits light or causes a change in wavelength of the incident light
- G01N21/66—Systems in which the material investigated is excited whereby it emits light or causes a change in wavelength of the incident light electrically excited, e.g. electroluminescence
-
- 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
- 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
- G01N21/95—Investigating the presence of flaws or contamination characterised by the material or shape of the object to be examined
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/25—Fusion techniques
-
- 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
- G06N3/084—Backpropagation, e.g. using gradient descent
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/0002—Inspection of images, e.g. flaw detection
- G06T7/0004—Industrial image inspection
- G06T7/0008—Industrial image inspection checking presence/absence
-
- 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
- G01N2021/8883—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 involving the calculation of gauges, generating models
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20081—Training; Learning
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20084—Artificial neural networks [ANN]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30108—Industrial image inspection
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Life Sciences & Earth Sciences (AREA)
- Health & Medical Sciences (AREA)
- Artificial Intelligence (AREA)
- General Health & Medical Sciences (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Computation (AREA)
- General Engineering & Computer Science (AREA)
- Bioinformatics & Computational Biology (AREA)
- Evolutionary Biology (AREA)
- Chemical & Material Sciences (AREA)
- Pathology (AREA)
- Immunology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Biochemistry (AREA)
- Analytical Chemistry (AREA)
- Biomedical Technology (AREA)
- Software Systems (AREA)
- Mathematical Physics (AREA)
- Computing Systems (AREA)
- Molecular Biology (AREA)
- Computational Linguistics (AREA)
- Biophysics (AREA)
- Quality & Reliability (AREA)
- Signal Processing (AREA)
- Nuclear Medicine, Radiotherapy & Molecular Imaging (AREA)
- Image Analysis (AREA)
Abstract
Aiming at the defects of the crack defect recognition technology of the photovoltaic cell EL image, the invention firstly adopts a near infrared camera to collect the photovoltaic cell EL image, and the pixels are 1024 multiplied by 1024; processing the acquired EL image of four fifths into an image block of 128×128 pixels, making an artificial label on the image block, including crack and non-crack labels, and constructing a training set; the remaining one-fifth of the EL images were taken as test sets. The invention uses convolutional neural network for crack recognition, and the input of the network is an EL image block of 128×128 pixels; and when the test is carried out, inputting an EL image with 1024 multiplied by 1024 pixels, calculating the position of the output crack through a neural network in a sliding window mode, marking the position with a frame, calculating the information entropy H value of the marked crack, and judging the authenticity of the marked object belonging to the crack.
Description
Technical Field
The invention relates to the technical field of crack defect identification of photovoltaic cell EL images, in particular to a crack identification algorithm based on a convolutional neural network and information entropy data fusion strategy.
Background
Photovoltaic power generation has become one of the most popular renewable energy sources in the world due to the characteristics of cleanliness, safety, high efficiency and the like. Photovoltaic power generation refers to a process of converting solar energy into electric energy by using photovoltaic cells. However, crack defects are inevitably caused during manufacturing, transportation and installation due to the vulnerability of the photovoltaic cell crystalline material. The existence of cracks can influence the power generation efficiency of the photovoltaic cell, reduce the service life and even influence the safe operation of the whole photovoltaic system. At present, the automatic crack detection technology for the EL image of the photovoltaic cell is not perfect, and a manual detection method is mostly adopted. The method has the characteristics of time consumption, low efficiency, subjectivity, high cost and the like, and further can influence the product quality of the photovoltaic cell. The invention adopts the convolutional neural network as the main method for recognizing the EL defects of the photovoltaic cells so as to realize quality inspection automation, and has important value and significance.
Disclosure of Invention
Aiming at the defects of the crack defect recognition technology of the photovoltaic cell EL image, the invention aims to solve the technical problem of inventing a crack recognition algorithm based on a convolutional neural network and information entropy data fusion strategy. The algorithm firstly adopts a near infrared camera to collect the EL image of a photovoltaic cell, and the pixels are 1024 multiplied by 1024; processing the acquired EL image of four fifths into an image block of 128×128 pixels, making an artificial label on the image block, including crack and non-crack labels, and constructing a training set; the remaining one-fifth EL image was taken as the test set. And constructing a convolutional neural network, setting proper parameters, and inputting a training set into the convolutional neural network for training to obtain model training weights. And when the test is carried out, inputting an EL image with 1024 multiplied by 1024 pixels, calculating the position of the output crack through a neural network in a sliding window mode, marking the position with a frame, calculating the information entropy H value of the marked crack, and judging the authenticity of the marked object belonging to the crack. Discarding the decision if false detection is made, and retaining if the crack is true. The method can identify crack defects of the high-resolution photovoltaic cell EL image, can remove false detection by adopting an information entropy data fusion algorithm, improves the accuracy of defect identification, releases manpower, and ensures the production quality of the photovoltaic cell.
In order to reduce the false detection rate of crack identification of a photovoltaic cell EL image, the invention provides a crack identification algorithm of the photovoltaic cell EL image based on a convolution neural network and information entropy data fusion strategy, which comprises the following specific steps:
first, data set preparation
1-1 image acquisition: acquiring an EL image of a photovoltaic cell by adopting a near infrared camera, wherein the pixels are 1024 multiplied by 1024;
1-2 segmentation of images: selecting four fifths of the EL image of the photovoltaic cell on the basis of the step 1-1, and dividing the EL image into image blocks of 128×128 pixels;
1-3 training set preparation: based on the step 1-2, making an artificial label for the image block, dividing the image block into a crack image block and a non-crack image block, and obtaining the image block containing the label as a training set;
1-4 test set fabrication: and on the basis of the step 1-2, selecting one fifth of the EL images of the photovoltaic cells in the step 1-1 as a test set.
Second, constructing convolutional neural network for crack recognition
2-1 convolutional neural network structure: the designed convolutional neural network mainly comprises a plurality of convolutional layers, a pooling layer and a full-connection layer, wherein Batch Normalization layers are connected behind the convolutional layers, and two adjacent layers are connected by a Relu layer;
the training method of the 2-2 convolutional neural network comprises the following steps: inputting the training set into the convolutional neural network on the basis of the step 2-1, and updating parameters of the convolutional neural network by adopting a back propagation algorithm;
2-3 parameter initialization: initializing a weight value to be a normal distribution random number obeying a standard deviation equal to 0.01, and initializing a bias term to be a constant; determining a learning rate parameter LR, a batch processing parameter batch and a training step number step;
2-4 update parameters: inputting the training set into a convolutional neural network on the basis of the step 2-3, calculating the output of each middle layer and the output layer, and calculating the magnitude of a loss function between the output and the artificial tag; calculating the gradient of the parameter of each layer by using the loss function; and updating the parameters of each layer according to the learning rate and the gradient.
Third step, test image
3-1 convolutional neural network crack recognition: based on the step 2-4, testing the EL image of the photovoltaic cell by adopting a sliding window mode, and marking a crack area in the EL image of the photovoltaic cell with 1024×1024 pixels by using a target frame with 128×128 pixels;
3-2 obtaining a Group of crack image blocks: based on the step 3-1, moving eight pixels upwards, downwards, leftwards, rightwards, leftwards upwards, leftwards downwards, rightwards upwards and rightwards respectively by taking the image block marked as a crack in the test image as a center to obtain eight image blocks, and forming a Group with the marked crack image blocks, wherein the sliding mode is shown in figure 2;
3-3 calculating the value of the information entropy H: based on the step 3-2, obtaining the probability value of each image block belonging to the crack in the Group, giving the probability value by a convolutional neural network, and calculating the value of the information entropy H marked with the crack;
3-4 test results: on the basis of the step 3-3, the convolution neural network is judged according to the threshold discriminant of the information entropy H, the image blocks marked with the cracks are defined as the authenticity of the cracks, if true, the image blocks are reserved, and if false, the image blocks are discarded.
Further, in the step 2-4, the magnitude of the loss function of the convolutional neural network is the difference between the output of the convolutional neural network and the artificial tag of the training set, and the loss function used for training the convolutional neural network is a cross entropy loss function, and the formula is as follows:
wherein y is i And y i ' represents the label of the ith image and the prediction of the output, respectively, M is the number of images.
Further, in the step 3-3, the information entropy H is a determination of the authenticity of the image block with marked cracks belonging to the cracks, and the calculation formula is as follows:
wherein P (Ccrack|p 1 ,…p 9 ) The probability value for each image block in a Group belonging to a crack is given by a convolutional neural network. The meaning of the parameter T is a coefficient of a constraint value of the information entropy, and different H values can be obtained by setting different T values. 2278 cracked image blocks and 2439 non-cracked image blocks were tested, their H values were calculated by setting different T values, and when t=4.5 was set, the cracked and non-cracked H values could be distinguished, resulting in fig. 3.
Compared with the prior art, the invention has the beneficial effects that:
the invention adopts a convolutional neural network for recognizing the EL image cracks of the photovoltaic cell, the network input is an image block containing an artificial label, and the positions of the cracks are output and marked by using a marking frame after the calculation of the convolutional neural network. And judging the authenticity of the identification result of the crack area obtained by the convolutional neural network. And acquiring a Group of marked crack areas, calculating an information entropy H value, judging the authenticity of the crack according to the threshold value of H, and discarding if the crack is a false crack. Compared with the traditional defect detection method, the algorithm improves the traditional convolutional neural network crack defect detection method, and proposes a data fusion strategy combined with information entropy, so that false detection generated in detection can be reduced, and the crack recognition accuracy of the EL image of the photovoltaic cell is improved. The method has strong algorithm applicability and high recognition accuracy.
Drawings
FIG. 1 is a flowchart of an algorithm of the present invention.
FIG. 2 is a schematic diagram of the present invention for obtaining groups for marked crack targets.
FIG. 3 is a scatter plot of the information entropy H values for crack and non-crack targets according to the present invention. (a) An H-value profile of a crack, (b) an H-value profile of a non-crack. Where t=4.5.
Detailed Description
The invention is further described below with reference to the drawings and examples of implementation.
The crack recognition algorithm of the EL image of the photovoltaic cell designed by the invention is mainly applied to industrial sites, and the algorithm is fully suitable for site conditions, and comprises the following specific steps:
first, data set preparation
1-1 image acquisition: acquiring an EL image of a photovoltaic cell by adopting a near infrared camera, wherein the pixels are 1024 multiplied by 1024;
1-2 segmentation of images: selecting four fifths of the EL image of the photovoltaic cell on the basis of the step 1-1, and dividing the EL image into image blocks of 128×128 pixels;
1-3 training set preparation: based on the step 1-2, making an artificial label for the image block, dividing the image block into a crack image block and a non-crack image block, and obtaining the image block containing the label as a training set;
1-4 test set fabrication: and on the basis of the step 1-2, selecting one fifth of the EL images of the photovoltaic cells in the step 1-1 as a test set.
Second, constructing convolutional neural network for crack recognition
2-1 convolutional neural network structure: the designed convolutional neural network mainly comprises a plurality of convolutional layers, a pooling layer and a full-connection layer, wherein Batch Normalization layers are connected behind the convolutional layers, and two adjacent layers are connected by a Relu layer;
the training method of the 2-2 convolutional neural network comprises the following steps: inputting the training set into the convolutional neural network on the basis of the step 2-1, and updating parameters of the convolutional neural network by adopting a back propagation algorithm;
2-3 parameter initialization: initializing a weight value to be a normal distribution random number obeying a standard deviation equal to 0.01, and initializing a bias term to be a constant; determining a learning rate parameter LR, a batch processing parameter batch and a training step number step;
2-4 update parameters: inputting the training set into a convolutional neural network on the basis of the step 2-3, calculating the output of each middle layer and the output layer, and calculating the magnitude of a loss function between the output and the artificial tag; calculating the gradient of the parameter of each layer by using the loss function; and updating the parameters of each layer according to the learning rate and the gradient.
Third step, test image
3-1 convolutional neural network crack recognition: based on the step 2-4, testing the EL image of the photovoltaic cell by adopting a sliding window mode, and marking a crack area in the EL image of the photovoltaic cell with 1024×1024 pixels by using a target frame with 128×128 pixels;
3-2 obtaining a Group of crack image blocks: based on the step 3-1, moving eight pixels upwards, downwards, leftwards, rightwards, leftwards upwards, leftwards downwards, rightwards upwards and rightwards respectively by taking the image block marked as a crack in the test image as a center to obtain eight image blocks, and forming a Group with the marked crack image blocks, wherein the sliding mode is shown in figure 2;
3-3 calculating the value of the information entropy H: based on the step 3-2, obtaining the probability value of each image block belonging to the crack in the Group, giving the probability value by a convolutional neural network, and calculating the value of the information entropy H marked with the crack;
3-4 test results: on the basis of the step 3-3, the convolution neural network is judged according to the threshold discriminant of the information entropy H, the image blocks marked with the cracks are defined as the authenticity of the cracks, if true, the image blocks are reserved, and if false, the image blocks are discarded.
The method is further characterized in that the pixel resolution of the EL image of the photovoltaic cell is 1024 multiplied by 1024, and the background and crack forms are obviously different.
The method is further characterized in that the design of the convolutional neural network is carried out, the input is connected with a convolutional layer, the convolutional layer adopts a convolutional kernel size of kernel=3, the step size of stride=1, and the edge filling is zero; the core size of the pooling layer comprises kernel=3 and kernel=2, step size stride=1, edge padding is zero, and the number of convolution layers and pooling layers in the network design is not equal.
The method is further characterized in that a convolution neural network is designed, batch Normalization layers are used between two adjacent layers, and an input layer inputs batch picture data x 1…m Conversion to BN (x) 1… m) has the following formula.
Wherein E [ x ] i ]Is the mean value Var [ x ] i ]Is the variance estimate and e is a constant added to the variance to ensure that the value is stable. To ensure that the content of the output representation does not change after normalization, parameters γ and β are introduced, which together with the original model parameters learn and recover the distribution of features learned from the original network, the formula is as follows.
β (k) ＝E[x (k) ]
The forward propagation process of the BN layer, the output is:
the method is further characterized in that when the threshold value of the information entropy is T=4.5, the H values of cracks and non-cracks can be well separated, and the threshold value theta= -1.5 is generated, and the formula is as follows.
The method is further characterized in that the performance of the image is evaluated, the invention adopts Precision, recall and F-measure to measure the performance, and the definition of the Precision, recall and F-measure are as follows:
wherein TP represents true positive, i.e. the image marked as defective is correctly identified; FP represents false positives, i.e. images marked as good are incorrectly identified as defective; FN indicates false negatives, i.e., images marked as defective are erroneously identified as non-defective.
The type of the invention is not described and is applicable to the prior art.
Claims (3)
1. The crack identification algorithm based on the convolutional neural network and the information entropy data fusion strategy is characterized by comprising the following steps of:
first, data set preparation
1-1 image acquisition: acquiring an EL image of a photovoltaic cell by adopting a near infrared camera, wherein the pixels are 1024 multiplied by 1024;
1-2 segmentation of images: selecting four fifths of the EL image of the photovoltaic cell on the basis of the step 1-1, and dividing the EL image into image blocks of 128×128 pixels;
1-3 training set preparation: based on the step 1-2, making an artificial label for the image block, dividing the image block into a crack image block and a non-crack image block, and obtaining the image block containing the label as a training set;
1-4 test set fabrication: selecting one fifth of the EL images of the photovoltaic cells in the step 1-1 as a test set on the basis of the step 1-2;
second, constructing convolutional neural network for crack recognition
2-1 convolutional neural network structure: the designed convolutional neural network mainly comprises a plurality of convolutional layers, a pooling layer and a full-connection layer, wherein Batch Normalization layers are connected behind the convolutional layers, and two adjacent layers are connected by a Relu layer;
the training method of the 2-2 convolutional neural network comprises the following steps: inputting the training set into the convolutional neural network on the basis of the step 2-1, and updating parameters of the convolutional neural network by adopting a back propagation algorithm;
2-3 parameter initialization: initializing a weight value to be a normal distribution random number obeying a standard deviation equal to 0.01, and initializing a bias term to be a constant; determining a learning rate parameter LR, a batch processing parameter batch and a training step number step;
2-4 update parameters: inputting the training set into a convolutional neural network on the basis of the step 2-3, calculating the output of each middle layer and the output layer, and calculating the magnitude of a loss function between the output and the artificial tag; calculating the gradient of the parameter of each layer by using the loss function; updating parameters of each layer according to the learning rate and the gradient;
third step, test image
3-1 convolutional neural network crack recognition: based on the step 2-4, testing the EL image of the photovoltaic cell by adopting a sliding window mode, and marking a crack area in the EL image of the photovoltaic cell with 1024×1024 pixels by using a target frame with 128×128 pixels;
3-2 obtaining a Group of crack image blocks: on the basis of the step 3-1, taking the image block marked as a crack in the test image as a center, respectively moving eight pixels upwards, downwards, leftwards, rightwards, upwards leftwards, downwards leftwards, upwards rightwards and downwards rightwards to obtain eight image blocks, and forming a Group with the marked crack image blocks;
3-3 calculating the value of the information entropy H: based on the step 3-2, obtaining the probability value of each image block belonging to the crack in the Group, giving the probability value by a convolutional neural network, and calculating the value of the information entropy H marked with the crack;
the information entropy H is used for judging the authenticity of the marked crack image block belonging to the crack, and the calculation formula is as follows:
wherein P (Ccrack|p 1 ,…p 9 ) For the probability value of each image block belonging to a crack in a Group, given by a convolutional neural network, the meaning of a parameter T is a coefficient of a constraint value of information entropy, different H values can be obtained by setting different T values, 2278 crack image blocks and 2439 non-crack image blocks are tested, the H values of the crack image blocks and the 2439 non-crack image blocks are calculated, and when T=4.5 is set, the crack and the non-crack H values can be distinguished;
3-4 classification of test results: on the basis of the step 3-3, the convolution neural network is judged according to the threshold discriminant of the information entropy H, the image blocks marked with the cracks are defined as the authenticity of the cracks, if true, the image blocks are reserved, and if false, the image blocks are discarded.
2. The crack recognition algorithm based on the convolutional neural network and information entropy data fusion strategy according to claim 1, wherein the threshold value of the information entropy in the step 3-4 is that when t=4.5, the H values of the crack and the non-crack can be well separated, and a threshold value θ= -1.5 is generated, and the formula is as follows:
3. the crack recognition algorithm based on the fusion strategy of the convolutional neural network and the information entropy data according to claim 1, wherein the output of the test results in the step 3-4 is characterized in that the EL image of the photovoltaic cell is input into the convolutional neural network, the crack recognition is carried out on the EL image in a sliding window mode, the test results of the convolutional neural network are marked with crack areas by using a frame of 128×128, the authenticity of marked cracks is judged again by adopting the fusion strategy of the information entropy data, if true, the authenticity is reserved, and if false, the authenticity is discarded.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010094093.4A CN111291814B (en) | 2020-02-15 | 2020-02-15 | Crack identification algorithm based on convolutional neural network and information entropy data fusion strategy |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010094093.4A CN111291814B (en) | 2020-02-15 | 2020-02-15 | Crack identification algorithm based on convolutional neural network and information entropy data fusion strategy |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN111291814A CN111291814A (en) | 2020-06-16 |
| CN111291814B true CN111291814B (en) | 2023-06-02 |
Family
ID=71024502
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010094093.4A Active CN111291814B (en) | 2020-02-15 | 2020-02-15 | Crack identification algorithm based on convolutional neural network and information entropy data fusion strategy |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111291814B (en) |
Families Citing this family (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111968127B (en) * | 2020-07-06 | 2021-08-27 | 中国科学院计算技术研究所 | Cancer focus area identification method and system based on full-section pathological image |
| KR102336110B1 (en) * | 2020-08-28 | 2021-12-06 | 박재완 | Method for fine defects Inspection of Leather using Deep Learning Model |
| DE102021118876A1 (en) | 2021-07-21 | 2023-01-26 | Deutsches Zentrum für Luft- und Raumfahrt e.V. | Method for detecting defects and method and system for laying down fiber material |
| CN113781448B (en) * | 2021-09-14 | 2024-01-23 | 国电四子王旗光伏发电有限公司 | Intelligent defect identification method for photovoltaic power station assembly based on infrared image analysis |
| CN117876359B (en) * | 2024-03-04 | 2024-06-14 | 金品计算机科技(天津)有限公司 | Rubber product defect detection method, device, equipment and medium |
Citations (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108717512A (en) * | 2018-05-16 | 2018-10-30 | 中国人民解放军陆军炮兵防空兵学院郑州校区 | A kind of malicious code sorting technique based on convolutional neural networks |
| CN109376792A (en) * | 2018-11-07 | 2019-02-22 | 河北工业大学 | Photovoltaic cell open defect classification method based on multichannel residual error neural network |
| CN109389043A (en) * | 2018-09-10 | 2019-02-26 | 中国人民解放军陆军工程大学 | A kind of crowd density estimation method of unmanned plane picture |
| CN109635784A (en) * | 2019-01-10 | 2019-04-16 | 重庆邮电大学 | Traffic sign recognition method based on improved convolutional neural networks |
| CN109873610A (en) * | 2019-03-19 | 2019-06-11 | 福州大学 | Diagnosing failure of photovoltaic array method based on IV characteristic and depth residual error network |
| CN109871883A (en) * | 2019-01-24 | 2019-06-11 | 北京市商汤科技开发有限公司 | Neural network training method and device, electronic equipment and storage medium |
| CN109918752A (en) * | 2019-02-26 | 2019-06-21 | 华南理工大学 | Mechanical failure diagnostic method, equipment and medium based on migration convolutional neural networks |
| CN109934203A (en) * | 2019-03-25 | 2019-06-25 | 南京大学 | A kind of cost-sensitive increment type face identification method based on comentropy selection |
| CN110119811A (en) * | 2019-05-15 | 2019-08-13 | 电科瑞达(成都)科技有限公司 | A kind of convolution kernel method of cutting out based on entropy significance criteria model |
| CN110197514A (en) * | 2019-06-13 | 2019-09-03 | 南京农业大学 | A kind of mushroom phenotype image generating method based on production confrontation network |
| CN110610475A (en) * | 2019-07-07 | 2019-12-24 | 河北工业大学 | Visual defect detection method of deep convolutional neural network |
-
2020
- 2020-02-15 CN CN202010094093.4A patent/CN111291814B/en active Active
Patent Citations (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108717512A (en) * | 2018-05-16 | 2018-10-30 | 中国人民解放军陆军炮兵防空兵学院郑州校区 | A kind of malicious code sorting technique based on convolutional neural networks |
| CN109389043A (en) * | 2018-09-10 | 2019-02-26 | 中国人民解放军陆军工程大学 | A kind of crowd density estimation method of unmanned plane picture |
| CN109376792A (en) * | 2018-11-07 | 2019-02-22 | 河北工业大学 | Photovoltaic cell open defect classification method based on multichannel residual error neural network |
| CN109635784A (en) * | 2019-01-10 | 2019-04-16 | 重庆邮电大学 | Traffic sign recognition method based on improved convolutional neural networks |
| CN109871883A (en) * | 2019-01-24 | 2019-06-11 | 北京市商汤科技开发有限公司 | Neural network training method and device, electronic equipment and storage medium |
| CN109918752A (en) * | 2019-02-26 | 2019-06-21 | 华南理工大学 | Mechanical failure diagnostic method, equipment and medium based on migration convolutional neural networks |
| CN109873610A (en) * | 2019-03-19 | 2019-06-11 | 福州大学 | Diagnosing failure of photovoltaic array method based on IV characteristic and depth residual error network |
| CN109934203A (en) * | 2019-03-25 | 2019-06-25 | 南京大学 | A kind of cost-sensitive increment type face identification method based on comentropy selection |
| CN110119811A (en) * | 2019-05-15 | 2019-08-13 | 电科瑞达(成都)科技有限公司 | A kind of convolution kernel method of cutting out based on entropy significance criteria model |
| CN110197514A (en) * | 2019-06-13 | 2019-09-03 | 南京农业大学 | A kind of mushroom phenotype image generating method based on production confrontation network |
| CN110610475A (en) * | 2019-07-07 | 2019-12-24 | 河北工业大学 | Visual defect detection method of deep convolutional neural network |
Non-Patent Citations (2)
| Title |
|---|
| An Entropy-based Pruning Method for CNN Compression;Jian-Hao Luo 等;《arXiv》;20170619;正文第1-10页 * |
| 基于信息熵和深度学习的无参考图像质量评价方法研究;卢 鹏 等;《计算机应用研究》;20181130;摘要，第1-2节 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN111291814A (en) | 2020-06-16 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111291814B (en) | Crack identification algorithm based on convolutional neural network and information entropy data fusion strategy | |
| CN109376792B (en) | Photovoltaic cell appearance defect classification method based on multi-channel residual error neural network | |
| CN108961235B (en) | Defective insulator identification method based on YOLOv3 network and particle filter algorithm | |
| WO2023155069A1 (en) | Deep-learning-based surface defect detection method for mobile phone battery | |
| CN110689011A (en) | Solar cell panel defect detection method of multi-scale combined convolution neural network | |
| CN114627383B (en) | Small sample defect detection method based on metric learning | |
| CN110610475A (en) | Visual defect detection method of deep convolutional neural network | |
| CN112508090A (en) | External package defect detection method | |
| CN113837994B (en) | Photovoltaic panel defect diagnosis method based on edge detection convolutional neural network | |
| CN113781466B (en) | Solar panel defect identification method based on multi-scale attention-dense network | |
| CN114581782A (en) | Fine defect detection method based on coarse-to-fine detection strategy | |
| CN112258490A (en) | Low-emissivity coating intelligent damage detection method based on optical and infrared image fusion | |
| CN115861190A (en) | Comparison learning-based unsupervised defect detection method for photovoltaic module | |
| CN112258459A (en) | Lithium battery defect detection method based on integrated deep learning | |
| CN114612406A (en) | Photovoltaic panel defect detection method based on visible light and infrared vision | |
| CN113096085A (en) | Container surface damage detection method based on two-stage convolutional neural network | |
| CN115375672A (en) | Coal gangue detection method based on improved YOLOv4 algorithm | |
| CN114897857A (en) | Solar cell defect detection method based on light neural network | |
| CN112381794B (en) | Printing defect detection method based on deep convolution generation network | |
| CN113962929A (en) | Photovoltaic cell assembly defect detection method and system and photovoltaic cell assembly production line | |
| CN115830302B (en) | Multi-scale feature extraction fusion power distribution network equipment positioning identification method | |
| CN116994161A (en) | Insulator defect detection method based on improved YOLOv5 | |
| CN116152194A (en) | Object defect detection method, system, equipment and medium | |
| CN114863211A (en) | Magnetic shoe defect detection and segmentation method based on deep learning | |
| CN116188877B (en) | Method and system for detecting and classifying unknown wafer defect categories |
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 |