CN114897850A - Supervised Dropout method and equipment for detecting surface defects of strip steel - Google Patents
Supervised Dropout method and equipment for detecting surface defects of strip steel Download PDFInfo
- Publication number
- CN114897850A CN114897850A CN202210558665.9A CN202210558665A CN114897850A CN 114897850 A CN114897850 A CN 114897850A CN 202210558665 A CN202210558665 A CN 202210558665A CN 114897850 A CN114897850 A CN 114897850A
- Authority
- CN
- China
- Prior art keywords
- strip steel
- supervised
- surface defects
- dropout
- strip
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
- 230000007547 defect Effects 0.000 title claims abstract description 159
- 229910000831 Steel Inorganic materials 0.000 title claims abstract description 155
- 239000010959 steel Substances 0.000 title claims abstract description 155
- 238000000034 method Methods 0.000 title claims abstract description 84
- 238000012549 training Methods 0.000 claims abstract description 33
- 238000012360 testing method Methods 0.000 claims abstract description 25
- 210000002569 neuron Anatomy 0.000 claims description 26
- 239000011159 matrix material Substances 0.000 claims description 24
- 238000000605 extraction Methods 0.000 claims description 16
- 230000006870 function Effects 0.000 claims description 14
- 238000009826 distribution Methods 0.000 claims description 13
- 230000002068 genetic effect Effects 0.000 claims description 13
- 230000003044 adaptive effect Effects 0.000 claims description 8
- 238000005457 optimization Methods 0.000 claims description 8
- 238000004891 communication Methods 0.000 claims description 7
- 238000005070 sampling Methods 0.000 claims description 7
- 238000003860 storage Methods 0.000 claims description 6
- 238000005520 cutting process Methods 0.000 claims description 5
- 230000035772 mutation Effects 0.000 claims description 5
- 230000008859 change Effects 0.000 claims description 4
- 238000006073 displacement reaction Methods 0.000 claims description 4
- 238000001514 detection method Methods 0.000 abstract description 6
- 238000013528 artificial neural network Methods 0.000 abstract description 5
- 238000013473 artificial intelligence Methods 0.000 abstract description 4
- 238000013135 deep learning Methods 0.000 abstract description 4
- 238000010586 diagram Methods 0.000 description 16
- 238000004519 manufacturing process Methods 0.000 description 6
- 238000007781 pre-processing Methods 0.000 description 3
- 230000008569 process Effects 0.000 description 3
- XEEYBQQBJWHFJM-UHFFFAOYSA-N Iron Chemical compound [Fe] XEEYBQQBJWHFJM-UHFFFAOYSA-N 0.000 description 2
- 230000007797 corrosion Effects 0.000 description 2
- 238000005260 corrosion Methods 0.000 description 2
- 230000003287 optical effect Effects 0.000 description 2
- 238000005096 rolling process Methods 0.000 description 2
- 241000519995 Stachys sylvatica Species 0.000 description 1
- 238000004590 computer program Methods 0.000 description 1
- 230000008094 contradictory effect Effects 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 230000002779 inactivation Effects 0.000 description 1
- 238000009776 industrial production Methods 0.000 description 1
- 229910052742 iron Inorganic materials 0.000 description 1
- 239000000203 mixture Substances 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000011176 pooling Methods 0.000 description 1
- 238000002203 pretreatment Methods 0.000 description 1
- 238000012545 processing Methods 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 238000003466 welding Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/0002—Inspection of images, e.g. flaw detection
- G06T7/0004—Industrial image inspection
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/764—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using classification, e.g. of video objects
- G06V10/765—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using classification, e.g. of video objects using rules for classification or partitioning the feature space
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/77—Processing image or video features in feature spaces; using data integration or data reduction, e.g. principal component analysis [PCA] or independent component analysis [ICA] or self-organising maps [SOM]; Blind source separation
- G06V10/7715—Feature extraction, e.g. by transforming the feature space, e.g. multi-dimensional scaling [MDS]; Mappings, e.g. subspace methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/77—Processing image or video features in feature spaces; using data integration or data reduction, e.g. principal component analysis [PCA] or independent component analysis [ICA] or self-organising maps [SOM]; Blind source separation
- G06V10/774—Generating sets of training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/82—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using neural networks
-
- 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
- G06T2207/30136—Metal
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02P—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN THE PRODUCTION OR PROCESSING OF GOODS
- Y02P90/00—Enabling technologies with a potential contribution to greenhouse gas [GHG] emissions mitigation
- Y02P90/30—Computing systems specially adapted for manufacturing
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Evolutionary Computation (AREA)
- General Physics & Mathematics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Artificial Intelligence (AREA)
- Health & Medical Sciences (AREA)
- Software Systems (AREA)
- Computing Systems (AREA)
- General Health & Medical Sciences (AREA)
- Databases & Information Systems (AREA)
- Multimedia (AREA)
- Medical Informatics (AREA)
- Biophysics (AREA)
- Mathematical Physics (AREA)
- Data Mining & Analysis (AREA)
- Life Sciences & Earth Sciences (AREA)
- Computational Linguistics (AREA)
- Molecular Biology (AREA)
- General Engineering & Computer Science (AREA)
- Biomedical Technology (AREA)
- Quality & Reliability (AREA)
- Image Analysis (AREA)
- Investigating Materials By The Use Of Optical Means Adapted For Particular Applications (AREA)
Abstract
The invention provides a supervised Dropout method and equipment for detecting surface defects of strip steel. The method comprises the following steps: step 1 to step 10. The method effectively solves the overfitting problem caused by unbalanced small sample training data sets when the deep learning network is applied to the strip steel surface quality defect detection, reduces the cost and threshold of artificial intelligence application in the field, guides the training of the deep neural network completely according to the result of network prediction classification, enables the network model to have better utilization rate on the trained strip steel defect data, and can obviously improve the test accuracy under the same prediction time.
Description
Technical Field
The embodiment of the invention relates to the technical field of image classification, in particular to a supervised Dropout method and equipment for detecting surface defects of strip steel.
Background
In the production process of strip steel, surface defects such as white spots, holes, roll marks, scratches, corrosion, splicing, black spots, edge cracks and the like are inevitably generated under the influence of factors such as rolling equipment, production process and the like. The surface defects of the strip steel not only can influence the appearance of a strip steel product and weaken the performance of the product such as iron loss characteristic, fatigue resistance, corrosion resistance and the like, but also can cause serious production accidents such as strip steel breakage, surfacing welding, production line shutdown and the like, and can generate immeasurable economic and social influences on production enterprises. Therefore, the defects in the rolling process are very necessary to be found in time, and the production process of the whole strip steel can be improved in a targeted manner through the detected strip steel defect data, so that the method has very important significance for improving the quality of the strip steel. The related strip steel defect detection method is difficult to obtain a good detection effect in the face of complex and variable actual industrial production conditions. For example, there are two major drawbacks to the standard Dropout, one of which is that the inactivation of neurons is randomly selected, lacking in pertinence; the other is that the performance of the neural network is continuously enhanced, and the discarding rate is fixed. Therefore, it is an urgent technical problem to be solved in the art to develop a supervised Dropout method and apparatus for detecting surface defects of strip steel, which can effectively overcome the above-mentioned defects in the related art.
Disclosure of Invention
Aiming at the problems in the prior art, the embodiment of the invention provides a supervised Dropout method and equipment for detecting the surface defects of the strip steel.
In a first aspect, an embodiment of the present invention provides a supervised Dropout method for detecting surface defects of a strip steel, including: step 1: collecting a strip steel surface defect image set; step 2: according to the collected data distribution of the strip steel surface defect image set, carrying out gradient data enhancement on the strip steel defect image set; and 3, step 3: the method comprises the steps of equalizing a strip steel surface defect image set after gradient data enhancement by adopting a random down-sampling mode; and 4, step 4: dividing the equalized strip steel surface defect image set into a training set and a testing set according to a preset proportion; and 5: sending the training set of the strip steel surface defect image into a network model; step 6: extracting the depth characteristics of the strip steel surface defect image by using a characteristic extraction layer of a network model; and 7: sending the extracted depth characteristics of the surface defects of the strip steel into a supervised Dropout to obtain an expected neuron mask matrix; and 8: inputting an expected neuron mask matrix into a classification network layer, completing forward propagation and backward propagation of a network to update parameters of a network model, and sequentially and circularly completing the training of the model; and step 9: inputting the test set image of the strip steel surface defect into a trained network model, and extracting the depth characteristic of the strip steel surface defect through a characteristic extraction layer; step 10: and sending the extracted depth characteristics of the surface defects of the strip steel into a classification layer, and outputting defect categories corresponding to the test images.
On the basis of the content of the embodiment of the method, the supervised Dropout method for detecting the surface defects of the strip steel provided by the embodiment of the invention performs gradient data enhancement on a strip steel defect data set, and comprises the following steps: and performing gradient data enhancement on the strip steel defect data set by adopting horizontal overturning, vertical overturning, horizontal and vertical overturning, brightness change, Gaussian noise, displacement and random cutting.
On the basis of the content of the embodiment of the method, the supervised Dropout method for detecting the surface defects of the strip steel provided by the embodiment of the invention comprises the following preset proportions: 4:1, 3:1, 5:1, 3.5:1, or 4.5: 1.
On the basis of the content of the embodiment of the method, the supervised Dropout method for detecting the surface defects of the strip steel provided by the embodiment of the invention comprises the following steps of: AlexNet, VGG, ResNet or EfficientNet network models.
On the basis of the content of the embodiment of the method, the supervised Dropout method for detecting the surface defects of the strip steel provided by the embodiment of the invention comprises the following specific steps of step 7: step 701: randomly generating n neuron mask matrixes according to a given discarding rate to serve as an initial population during genetic algorithm optimization, and replacing a first individual of the initial population with a last expected mask matrix; step 702: sending each neuron mask matrix and the extracted depth features into a classification layer together to obtain a corresponding classification result; step 703: calculating a cross entropy loss function value corresponding to each classification result, and taking the cross entropy loss function value as an adaptive value optimized by a genetic algorithm; step 704: comparing the fitness corresponding to the historical optimal individual with the fitness value of each individual of the population, and if the maximum fitness value of the population individual is less than or equal to the fitness value of the historical optimal individual, keeping the historical optimal individual unchanged; step 705: calculating the selection probability of each individual in the population according to the adaptive value; step 706: selecting individuals who are inherited to the next generation in a roulette manner according to the selection probability; step 707: performing cross operation on the selected individuals inherited to the next generation according to a first preset probability, and performing mutation operation according to a second preset probability; step 708: taking the population subjected to the cross and variation operation as a new population, and replacing the history optimal individual with a first individual of the new population; step 709: returning to the step 702, and circulating in sequence until the optimal individual is continuously updated for 5 times, and stopping genetic algorithm optimization; step 710: and outputting the history optimal individual as an expected neuron mask matrix.
On the basis of the content of the above method embodiment, the supervised Dropout method for detecting surface defects of strip steel provided in the embodiment of the present invention further includes, after the history optimal individual remains unchanged: and if the maximum fitness value of the population individuals is larger than the fitness value of the historical optimal individual, replacing the individual corresponding to the maximum fitness value with the historical optimal individual.
On the basis of the content of the embodiment of the method, in the supervised Dropout method for detecting the surface defects of the strip steel, the first preset probability is a preset probability with a probability value of 60%, and the second preset probability is a preset probability with a probability value of 10%.
In a second aspect, an embodiment of the present invention provides a supervised Dropout apparatus for detecting surface defects of a steel strip, including: a first master module, configured to implement step 1: collecting a strip steel surface defect image set; a second master module, configured to implement step 2: according to the collected data distribution of the strip steel surface defect image set, carrying out gradient data enhancement on the strip steel defect image set; a third main module, configured to implement step 3: the method comprises the steps of equalizing a strip steel surface defect image set after gradient data enhancement by adopting a random down-sampling mode; a fourth master module, configured to implement step 4: dividing the equalized strip steel surface defect image set into a training set and a testing set according to a preset proportion; a fifth master module, configured to implement step 5: sending the training set of the strip steel surface defect image into a network model; a sixth master module, configured to implement step 6: extracting the depth characteristics of the strip steel surface defect image by using a characteristic extraction layer of a network model; a seventh master module, configured to implement step 7: sending the extracted depth characteristics of the surface defects of the strip steel into a supervised Dropout to obtain an expected neuron mask matrix; an eighth master module, configured to implement step 8: inputting an expected neuron mask matrix into a classification network layer, completing forward propagation and backward propagation of a network to update parameters of a network model, and sequentially and circularly completing the training of the model; a ninth master module, configured to implement step 9: inputting the test set image of the strip steel surface defect into a trained network model, and extracting the depth characteristic of the strip steel surface defect through a characteristic extraction layer; a tenth main module, configured to implement step 10: and sending the extracted depth characteristics of the surface defects of the strip steel into a classification layer, and outputting defect categories corresponding to the test images.
In a third aspect, an embodiment of the present invention provides an electronic device, including:
at least one processor; and
at least one memory communicatively coupled to the processor, wherein:
the memory stores program instructions executable by the processor, and the processor calls the program instructions to execute the supervised Dropout method for detecting the surface defects of the strip steel provided by any one of the various implementation manners of the first aspect.
In a fourth aspect, an embodiment of the present invention provides a non-transitory computer-readable storage medium storing computer instructions for causing a computer to execute the supervised Dropout method for detecting surface defects of a steel strip provided in any one of the various implementation manners of the first aspect.
The supervised Dropout method and the supervised Dropout equipment for detecting the surface defects of the strip steel effectively solve the problem of overfitting caused by unbalanced small sample training data sets when the deep learning network is applied to the surface quality defect detection of the strip steel, reduce the cost and threshold of the artificial intelligence application field, and guide the training of the deep neural network completely according to the result of network prediction classification, so that the network model has better utilization rate on the trained strip steel defect data, and can obviously improve the test accuracy rate in the same prediction time.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, a brief description will be given below to the drawings required for the description of the embodiments or the prior art, and it is obvious that the drawings in the following description are some embodiments of the present invention, and for those skilled in the art, other drawings can be obtained according to these drawings without creative efforts.
FIG. 1 is a flowchart of a supervised Dropout method for detecting surface defects of strip steel according to an embodiment of the present invention;
FIG. 2 is a schematic structural diagram of a supervised Dropout apparatus for detecting surface defects of strip steel according to an embodiment of the present invention;
fig. 3 is a schematic physical structure diagram of an electronic device according to an embodiment of the present invention;
FIG. 4 is a defect diagram of a strip steel provided by an embodiment of the invention;
FIG. 5 is a block diagram of a gradient random enhancement provided by an embodiment of the present invention;
FIG. 6 is a diagram illustrating an original data distribution of a strip steel surface defect image set and a data distribution diagram after data preprocessing according to an embodiment of the present invention;
FIG. 7 is a graph illustrating accuracy curves of network models with supervision Dropot, standard Dropot and no Dropot on a test set of surface defects of a strip steel according to an embodiment of the present invention;
fig. 8 is a block diagram of application supervision Dropout according to an embodiment of the present invention;
fig. 9 is a block diagram of applying the supervision Dropout to the AlexNet network model according to the embodiment of the present invention.
Detailed Description
In order to make the objects, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are some, but not all, embodiments of the present invention. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention. In addition, technical features of various embodiments or individual embodiments provided by the present invention may be arbitrarily combined with each other to form a feasible technical solution, and such combination is not limited by the sequence of steps and/or the structural composition mode, but must be realized by a person skilled in the art, and when the technical solution combination is contradictory or cannot be realized, such a technical solution combination should not be considered to exist and is not within the protection scope of the present invention.
The embodiment of the invention provides a supervised Dropout method for detecting the surface defects of strip steel, and referring to a figure 1, the method comprises the following steps: step 1: collecting a strip steel surface defect image set; step 2: according to the collected data distribution of the strip steel surface defect image set, carrying out gradient data enhancement on the strip steel defect image set; and step 3: the method comprises the steps of equalizing a strip steel surface defect image set after gradient data enhancement by adopting a random down-sampling mode; and 4, step 4: dividing the equalized strip steel surface defect image set into a training set and a testing set according to a preset proportion; and 5: sending the training set of the strip steel surface defect image into a network model; step 6: extracting the depth characteristics of the strip steel surface defect image by using a characteristic extraction layer of a network model; and 7: sending the extracted depth characteristics of the surface defects of the strip steel into a supervised Dropout to obtain an expected neuron mask matrix; and 8: inputting an expected neuron mask matrix into a classification network layer, completing forward propagation and backward propagation of a network to update parameters of a network model, and sequentially and circularly completing the training of the model; and step 9: inputting the test set image of the strip steel surface defect into a trained network model, and extracting the depth characteristic of the strip steel surface defect through a characteristic extraction layer; step 10: and sending the extracted depth characteristics of the surface defects of the strip steel into a classification layer, and outputting defect categories corresponding to the test images.
Specifically, refer to fig. 4 and 8 for an image of a surface defect of the strip steel and a block diagram to which the supervision Dropout is applied, respectively. The method for detecting the surface quality defects of the strip steel comprises a pretreatment method of a strip steel surface defect image, a strip steel surface defect depth characteristic extraction method, a supervised Dropout guide model training method and a strip steel surface defect classification method.
Referring to fig. 5 and 6, a diagram of gradient data enhancement and a data distribution diagram of the collected original data distribution and the data distribution after data preprocessing of the 37-type strip steel surface defects are shown respectively. The preprocessing method of the strip steel surface defect image is used for carrying out data expansion and equalization operation on the collected 37 types of strip steel surface defect images. According to the data distribution characteristics of the 37 types of strip steel surface defects, gradient data enhancement is carried out on the strip steel surface defect data by adopting horizontal turning, vertical turning, horizontal and vertical turning, brightness, Gaussian noise, shifting and random cutting. And then, on the strip steel defect data set after data enhancement, carrying out equalization operation on the data set by using random down-sampling.
The strip steel surface defect depth feature extraction method is used for extracting the depth feature of a strip steel surface defect image, and the extracted depth feature is used for training a network model and detecting and identifying the strip steel surface defect. Currently, the deep feature extraction method is completed by a feature extraction network, and its main components are: convolutional layers, pooling layers, active layers, and the like.
The supervised Dropout guiding model training method is used for guiding the training of a network model, the loss value predicted by the final model is used as the individual fitness value of the genetic algorithm, the worst performance in the sub-network corresponding to the self-adaptive discarding rate is found out for training, the robustness of the neurons in the sub-network is improved, and the generalization performance of the final model is improved; and meanwhile, the maximum discarding rate corresponding to the optimal distribution of the training round is used as the initial discarding rate of the next training round, so that a data-driven self-adaptive discarding rate is obtained. A block diagram of the network architecture to which the supervision Dropout is applied with AlexNet as the network framework is shown in fig. 9. The band steel surface defect classification method is used for classifying the band steel surface defects by combining the extracted depth characteristics, and mainly comprises the following components: a full link layer and an active layer.
Based on the content of the foregoing method embodiment, as an optional embodiment, the supervised Dropout method for detecting surface defects of strip steel provided in the embodiment of the present invention performs gradient data enhancement on a strip steel defect data set, including: and performing gradient data enhancement on the strip steel defect data set by adopting horizontal overturning, vertical overturning, horizontal and vertical overturning, brightness change, Gaussian noise, displacement and random cutting.
Based on the content of the above method embodiment, as an optional embodiment, the supervised Dropout method for detecting surface defects of strip steel provided in the embodiment of the present invention includes: 4:1, 3:1, 5:1, 3.5:1, or 4.5: 1.
Based on the content of the foregoing method embodiment, as an optional embodiment, the supervised Dropout method for detecting surface defects of strip steel provided in the embodiment of the present invention includes: AlexNet, VGG, ResNet, or EfficientNet network models.
Based on the content of the foregoing method embodiment, as an optional embodiment, in the supervised Dropout method for detecting surface defects of strip steel provided in the embodiment of the present invention, the specific step of step 7 includes: step 701: randomly generating n neuron mask matrixes according to a given discarding rate to serve as an initial population during genetic algorithm optimization, and replacing a first individual of the initial population with a last expected mask matrix; step 702: sending each neuron mask matrix and the extracted depth features into a classification layer together to obtain a corresponding classification result; step 703: calculating a cross entropy loss function value corresponding to each classification result, and taking the cross entropy loss function value as an adaptive value optimized by a genetic algorithm; step 704: comparing the fitness corresponding to the historical optimal individual with the fitness value of each individual of the population, and if the maximum fitness value of the population individual is less than or equal to the fitness value of the historical optimal individual, keeping the historical optimal individual unchanged; step 705: calculating the selection probability of each individual in the population according to the adaptive value; step 706: selecting individuals who are inherited to the next generation in a roulette manner according to the selection probability; step 707: performing cross operation on individuals selected to be inherited to a next generation according to a first preset probability, and performing mutation operation according to a second preset probability; step 708: taking the population subjected to the cross and variation operation as a new population, and replacing the history optimal individual with a first individual of the new population; step 709: returning to the step 702, and circulating in sequence until the optimal individual is continuously updated for 5 times, and stopping genetic algorithm optimization; step 710: and outputting the history optimal individual as an expected neuron mask matrix.
Referring specifically to fig. 7, the accuracy curves of the network models with supervision Dropout, standard Dropout and no Dropout on the strip surface defect test set are shown. Through comparison of the accuracy curves, the accuracy of the supervised Dropout on the strip steel surface curve test set can be intuitively found to be obviously superior to that of other methods, and the effectiveness of the supervised Dropout method for detecting the strip steel surface defects provided by the embodiment of the invention is proved.
Based on the content of the foregoing method embodiment, as an optional embodiment, the supervised Dropout method for detecting surface defects of strip steel provided in the embodiment of the present invention further includes, after the history optimal individual remains unchanged: and if the maximum fitness value of the population individuals is larger than the fitness value of the historical optimal individual, replacing the individual corresponding to the maximum fitness value with the historical optimal individual.
Based on the content of the above method embodiment, as an optional embodiment, in the supervised Dropout method for detecting surface defects of strip steel provided in the embodiment of the present invention, the first preset probability is a preset probability with a probability value of 60%, and the second preset probability is a preset probability with a probability value of 10%.
The supervised Dropout method for detecting the surface defects of the strip steel effectively solves the problem of overfitting caused by unbalanced small sample training data sets when the deep learning network is applied to the surface quality defect detection of the strip steel, reduces the cost and threshold of artificial intelligence application in the field, guides the training of the deep neural network completely according to the result of network prediction classification, enables the network model to have better utilization rate of the trained strip steel defect data, and can obviously improve the test accuracy rate in the same prediction time.
The implementation basis of the various embodiments of the present invention is realized by programmed processing performed by a device having a processor function. Therefore, in engineering practice, the technical solutions and functions thereof of the embodiments of the present invention can be packaged into various modules. Based on the actual situation, on the basis of the above embodiments, the embodiment of the invention provides a supervised Dropout device for detecting the surface defects of the strip steel, which is used for executing the supervised Dropout method for detecting the surface defects of the strip steel in the above method embodiments. Referring to fig. 2, the apparatus includes: a first master module, configured to implement step 1: collecting a strip steel surface defect image set; a second master module, configured to implement step 2: according to the collected data distribution of the strip steel surface defect image set, carrying out gradient data enhancement on the strip steel defect image set; a third main module, configured to implement step 3: the method comprises the steps of equalizing a strip steel surface defect image set after gradient data enhancement by adopting a random down-sampling mode; a fourth master module, configured to implement step 4: dividing the equalized strip steel surface defect image set into a training set and a testing set according to a preset proportion; a fifth master module, configured to implement step 5: sending the training set of the strip steel surface defect image into a network model; a sixth master module, configured to implement step 6: extracting the depth characteristics of the strip steel surface defect image by using a characteristic extraction layer of a network model; a seventh master module, configured to implement step 7: sending the extracted depth characteristics of the surface defects of the strip steel into a supervised Dropout to obtain an expected neuron mask matrix; an eighth master module, configured to implement step 8: inputting an expected neuron mask matrix into a classification network layer, completing forward propagation and backward propagation of a network to update parameters of a network model, and sequentially and circularly completing the training of the model; a ninth master module, configured to implement step 9: inputting the test set image of the strip steel surface defect into a trained network model, and extracting the depth characteristic of the strip steel surface defect through a characteristic extraction layer; a tenth main module, configured to implement step 10: and sending the extracted depth characteristics of the surface defects of the strip steel into a classification layer, and outputting defect categories corresponding to the test images.
The supervised Dropout device for detecting the surface defects of the strip steel, provided by the embodiment of the invention, adopts the modules in the graph 2, effectively solves the problem of overfitting caused by an unbalanced small sample training data set when a deep learning network is applied to the surface quality defect detection of the strip steel, reduces the cost and threshold of applying artificial intelligence to the field, guides the training of a deep neural network completely according to the result of network prediction classification, ensures that a network model has better utilization rate on the defect data of the trained strip steel, and can obviously improve the test accuracy rate in the same prediction time.
It should be noted that, the apparatus in the apparatus embodiment provided by the present invention may be used for implementing methods in other method embodiments provided by the present invention, except that corresponding function modules are provided, and the principle of the apparatus embodiment provided by the present invention is basically the same as that of the apparatus embodiment provided by the present invention, so long as a person skilled in the art obtains corresponding technical means by combining technical features on the basis of the apparatus embodiment described above, and obtains a technical solution formed by these technical means, on the premise of ensuring that the technical solution has practicability, the apparatus in the apparatus embodiment described above may be modified, so as to obtain a corresponding apparatus class embodiment, which is used for implementing methods in other method class embodiments. For example:
based on the content of the foregoing device embodiment, as an optional embodiment, the supervised Dropout device for detecting surface defects of strip steel provided in the embodiment of the present invention further includes: the first submodule is used for realizing gradient data enhancement on the strip steel defect data set, and comprises: and performing gradient data enhancement on the strip steel defect data set by adopting horizontal overturning, vertical overturning, horizontal and vertical overturning, brightness change, Gaussian noise, displacement and random cutting.
Based on the content of the above device embodiment, as an optional embodiment, the supervised Dropout device for detecting surface defects of strip steel provided in the embodiment of the present invention further includes: the second submodule is used for realizing the preset proportion and comprises: 4:1, 3:1, 5:1, 3.5:1, or 4.5: 1.
Based on the content of the above device embodiment, as an optional embodiment, the supervised Dropout device for detecting surface defects of strip steel provided in the embodiment of the present invention further includes: a third sub-module, configured to implement the network model, including: AlexNet, VGG, ResNet or EfficientNet network models.
Based on the content of the above device embodiment, as an optional embodiment, the supervised Dropout device for detecting surface defects of strip steel provided in the embodiment of the present invention further includes: a fourth sub-module, configured to implement the specific steps of step 7, including: step 701: randomly generating n neuron mask matrixes according to a given discarding rate to serve as an initial population during genetic algorithm optimization, and replacing a first individual of the initial population with a last expected mask matrix; step 702: sending each neuron mask matrix and the extracted depth features into a classification layer together to obtain a corresponding classification result; step 703: calculating a cross entropy loss function value corresponding to each classification result, and taking the cross entropy loss function value as an adaptive value optimized by a genetic algorithm; step 704: comparing the fitness corresponding to the historical optimal individual with the fitness value of each individual of the population, and if the maximum fitness value of the population individual is less than or equal to the fitness value of the historical optimal individual, keeping the historical optimal individual unchanged; step 705: calculating the selection probability of each individual in the population according to the adaptive value; step 706: selecting individuals who are inherited to the next generation in a roulette manner according to the selection probability; step 707: performing cross operation on individuals selected to be inherited to a next generation according to a first preset probability, and performing mutation operation according to a second preset probability; step 708: taking the population subjected to the cross and variation operation as a new population, and replacing the history optimal individual with a first individual of the new population; step 709: returning to the step 702, and circulating in sequence until the optimal individual is continuously updated for 5 times, and stopping genetic algorithm optimization; step 710: and outputting the history optimal individual as an expected neuron mask matrix.
Based on the content of the above device embodiment, as an optional embodiment, the supervised Dropout device for detecting surface defects of strip steel provided in the embodiment of the present invention further includes: a fifth sub-module, configured to implement that after the history optimal individual remains unchanged, the method further includes: and if the maximum fitness value of the population individuals is larger than the fitness value of the historical optimal individual, replacing the individual corresponding to the maximum fitness value with the historical optimal individual.
Based on the content of the above device embodiment, as an optional embodiment, the supervised Dropout device for detecting surface defects of strip steel provided in the embodiment of the present invention further includes: and the sixth submodule is used for realizing that the first preset probability is a preset probability with a probability value of 60 percent, and the second preset probability is a preset probability with a probability value of 10 percent.
The method of the embodiment of the invention is realized by depending on the electronic equipment, so that the related electronic equipment is necessarily introduced. To this end, an embodiment of the present invention provides an electronic apparatus, as shown in fig. 3, including: the system comprises at least one processor (processor), a communication Interface (communication Interface), at least one memory (memory) and a communication bus, wherein the at least one processor, the communication Interface and the at least one memory are communicated with each other through the communication bus. The at least one processor may invoke logic instructions in the at least one memory to perform all or a portion of the steps of the methods provided by the various method embodiments described above.
In addition, the logic instructions in the at least one memory may be implemented in software functional units and stored in a computer readable storage medium when sold or used as a stand-alone product. Based on such understanding, the technical solution of the present invention may be embodied in the form of a software product, which is stored in a storage medium and includes instructions for causing a computer device (which may be a personal computer, a server, or a network device) to execute all or part of the steps of the method according to the method embodiments of the present invention. And the aforementioned storage medium includes: a U-disk, a removable hard disk, a Read-Only Memory (ROM), a Random Access Memory (RAM), a magnetic disk or an optical disk, and other various media capable of storing program codes.
The above-described embodiments of the apparatus are merely illustrative, and the units described as separate parts may or may not be physically separate, and parts displayed as units may or may not be physical units, may be located in one place, or may be distributed on a plurality of network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of the present embodiment. One of ordinary skill in the art can understand and implement it without inventive effort.
Through the above description of the embodiments, those skilled in the art will clearly understand that each embodiment can be implemented by software plus a necessary general hardware platform, and certainly can also be implemented by hardware. With this understanding in mind, the above-described technical solutions may be embodied in the form of a software product, which can be stored in a computer-readable storage medium, such as ROM/RAM, magnetic disk, optical disk, etc., and includes instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to execute the methods described in the embodiments or some parts of the embodiments.
The flowchart and block diagrams in the figures illustrate the architecture, functionality, and operation of possible implementations of systems, methods and computer program products according to various embodiments of the present invention. Based on this recognition, each block in the flowchart or block diagrams may represent a module, a program segment, or a portion of code, which comprises one or more executable instructions for implementing the specified logical function(s). It should also be noted that, in some alternative implementations, the functions noted in the block may occur out of the order noted in the figures. For example, two blocks shown in succession may, in fact, be executed substantially concurrently, or the blocks may sometimes be executed in the reverse order, depending upon the functionality involved. It will also be noted that each block of the block diagrams and/or flowchart illustration, and combinations of blocks in the block diagrams and/or flowchart illustration, can be implemented by special purpose hardware-based systems which perform the specified functions or acts, or combinations of special purpose hardware and computer instructions.
It should be noted that the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising … …" does not exclude the presence of other identical elements in a process, method, article, or apparatus that comprises the element.
Finally, it should be noted that: the above examples are only intended to illustrate the technical solution of the present invention, but not to limit it; although the present invention has been described in detail with reference to the foregoing embodiments, it will be understood by those of ordinary skill in the art that: the technical solutions described in the foregoing embodiments may still be modified, or some technical features may be equivalently replaced; and such modifications or substitutions do not depart from the spirit and scope of the corresponding technical solutions of the embodiments of the present invention.
Claims (10)
1. A supervised Dropout method for detecting surface defects of strip steel is characterized by comprising the following steps: step 1: collecting a strip steel surface defect image set; step 2: according to the collected data distribution of the strip steel surface defect image set, carrying out gradient data enhancement on the strip steel defect image set; and step 3: the method comprises the steps of equalizing a strip steel surface defect image set after gradient data enhancement by adopting a random down-sampling mode; and 4, step 4: dividing the equalized strip steel surface defect image set into a training set and a testing set according to a preset proportion; and 5: sending the training set of the strip steel surface defect image into a network model; step 6: extracting the depth characteristics of the strip steel surface defect image by using a characteristic extraction layer of a network model; and 7: sending the extracted depth characteristics of the surface defects of the strip steel into a supervised Dropout to obtain an expected neuron mask matrix; and 8: inputting an expected neuron mask matrix into a classification network layer, completing forward propagation and backward propagation of a network to update parameters of a network model, and sequentially and circularly completing the training of the model; and step 9: inputting the test set image of the strip steel surface defect into a trained network model, and extracting the depth characteristic of the strip steel surface defect through a characteristic extraction layer; step 10: and sending the extracted depth characteristics of the surface defects of the strip steel into a classification layer, and outputting defect categories corresponding to the test images.
2. The supervised Dropout method for detecting surface defects of steel strip as recited in claim 1, wherein the performing gradient data enhancement on the steel strip defect data set comprises: and performing gradient data enhancement on the strip steel defect data set by adopting horizontal overturning, vertical overturning, horizontal and vertical overturning, brightness change, Gaussian noise, displacement and random cutting.
3. The supervised Dropout method for detecting surface defects of steel strip as recited in claim 2, wherein the preset ratio comprises: 4:1, 3:1, 5:1, 3.5:1, or 4.5: 1.
4. The supervised Dropout method for detecting surface defects of steel strip as recited in claim 3, wherein the network model comprises: AlexNet, VGG, ResNet or EfficientNet network models.
5. The supervised Dropout method for detecting surface defects of steel strip as recited in claim 4, wherein the specific steps of the step 7 include: step 701: randomly generating n neuron mask matrixes according to a given discarding rate to serve as an initial population during genetic algorithm optimization, and replacing a first individual of the initial population with a last expected mask matrix; step 702: sending each neuron mask matrix and the extracted depth features into a classification layer together to obtain a corresponding classification result; step 703: calculating a cross entropy loss function value corresponding to each classification result, and taking the cross entropy loss function value as an adaptive value optimized by a genetic algorithm; step 704: comparing the fitness corresponding to the historical optimal individual with the fitness value of each individual of the population, and if the maximum fitness value of the population individual is less than or equal to the fitness value of the historical optimal individual, keeping the historical optimal individual unchanged; step 705: calculating the selection probability of each individual in the population according to the adaptive value; step 706: selecting individuals who are inherited to the next generation in a roulette manner according to the selection probability; step 707: performing cross operation on the selected individuals inherited to the next generation according to a first preset probability, and performing mutation operation according to a second preset probability; step 708: taking the population subjected to the crossing and mutation operation as a new population, and replacing the historical optimal individual with a first individual of the new population; step 709: returning to the step 702, and circulating in sequence until the optimal individual is continuously updated for 5 times, and stopping genetic algorithm optimization; step 710: and outputting the history optimal individual as an expected neuron mask matrix.
6. The supervised Dropout method for detecting surface defects of steel strip as recited in claim 5, further comprising, after the historically optimal individual remains unchanged: and if the maximum fitness value of the population individuals is larger than the fitness value of the historical optimal individual, replacing the individual corresponding to the maximum fitness value with the historical optimal individual.
7. The supervised Dropout method for detecting surface defects of steel strip as recited in claim 6, wherein the first preset probability is a preset probability with a probability value of 60%, and the second preset probability is a preset probability with a probability value of 10%.
8. A supervised Dropout device for detecting surface defects of strip steel is characterized by comprising the following components: a first master module, configured to implement step 1: collecting a strip steel surface defect image set; a second master module, configured to implement step 2: according to the collected data distribution of the strip steel surface defect image set, carrying out gradient data enhancement on the strip steel defect image set; a third main module, configured to implement step 3: the method comprises the steps of equalizing a strip steel surface defect image set after gradient data enhancement by adopting a random down-sampling mode; a fourth master module, configured to implement step 4: dividing the equalized strip steel surface defect image set into a training set and a testing set according to a preset proportion; a fifth master module, configured to implement step 5: sending the training set of the strip steel surface defect image into a network model; a sixth master module, configured to implement step 6: extracting the depth characteristics of the strip steel surface defect image by using a characteristic extraction layer of a network model; a seventh master module, configured to implement step 7: sending the extracted depth characteristics of the surface defects of the strip steel into a supervised Dropout to obtain an expected neuron mask matrix; an eighth master module, configured to implement step 8: inputting an expected neuron mask matrix into a classification network layer, completing forward propagation and backward propagation of a network to update parameters of a network model, and sequentially and circularly completing the training of the model; a ninth master module, configured to implement step 9: inputting the test set image of the strip steel surface defect into a trained network model, and extracting the depth characteristic of the strip steel surface defect through a characteristic extraction layer; a tenth main module, configured to implement step 10: and sending the extracted depth characteristics of the surface defects of the strip steel into a classification layer, and outputting defect categories corresponding to the test images.
9. An electronic device, comprising:
at least one processor, at least one memory, and a communication interface; wherein,
the processor, the memory and the communication interface are communicated with each other;
the memory stores program instructions executable by the processor, the processor invoking the program instructions to perform the method of any of claims 1 to 7.
10. A non-transitory computer-readable storage medium storing computer instructions for causing a computer to perform the method of any one of claims 1 to 7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210558665.9A CN114897850B (en) | 2022-05-20 | 2022-05-20 | Method and equipment for detecting surface defect of strip steel by using supervised Dropout |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210558665.9A CN114897850B (en) | 2022-05-20 | 2022-05-20 | Method and equipment for detecting surface defect of strip steel by using supervised Dropout |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN114897850A true CN114897850A (en) | 2022-08-12 |
| CN114897850B CN114897850B (en) | 2024-08-06 |
Family
ID=82724523
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210558665.9A Active CN114897850B (en) | 2022-05-20 | 2022-05-20 | Method and equipment for detecting surface defect of strip steel by using supervised Dropout |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114897850B (en) |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| EP3540688A1 (en) * | 2018-03-14 | 2019-09-18 | OMRON Corporation | Defect inspection device, defect inspection method, and program |
| CN113034483A (en) * | 2021-04-07 | 2021-06-25 | 昆明理工大学 | Cigarette defect detection method based on deep migration learning |
| US20210279520A1 (en) * | 2020-03-09 | 2021-09-09 | Nanotronics Imaging, Inc. | Defect Detection System |
-
2022
- 2022-05-20 CN CN202210558665.9A patent/CN114897850B/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| EP3540688A1 (en) * | 2018-03-14 | 2019-09-18 | OMRON Corporation | Defect inspection device, defect inspection method, and program |
| US20210279520A1 (en) * | 2020-03-09 | 2021-09-09 | Nanotronics Imaging, Inc. | Defect Detection System |
| CN113034483A (en) * | 2021-04-07 | 2021-06-25 | 昆明理工大学 | Cigarette defect detection method based on deep migration learning |
Non-Patent Citations (1)
| Title |
|---|
| 周末;徐玲;杨梦宁;廖胜平;鄢萌;: "基于深度自编码网络的软件缺陷预测方法", 计算机工程与科学, no. 10, 15 October 2018 (2018-10-15) * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN114897850B (en) | 2024-08-06 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109584221B (en) | Abnormal image detection method based on supervised generation countermeasure network | |
| CN112949822B (en) | Low-perceptibility countermeasure sample composition method based on dual-attention mechanism | |
| CN108737406A (en) | A kind of detection method and system of abnormal flow data | |
| CN110503641A (en) | A kind of method and apparatus improving continuous casting billet face crack | |
| CN115222630A (en) | Image generation method, and training method and device of image denoising model | |
| CN113609482B (en) | Back door detection and restoration method and system for image classification model | |
| CN112861906A (en) | Method and device for detecting picture abnormity | |
| CN111723841A (en) | Text detection method and device, electronic equipment and storage medium | |
| CN117745680A (en) | Abnormality detection method and device based on large visual language model | |
| CN116755095A (en) | Short-term precipitation prediction method and device, terminal equipment and storage medium | |
| CN114580934A (en) | Early warning method for food detection data risk based on unsupervised anomaly detection | |
| CN116563250A (en) | Recovery type self-supervision defect detection method, device and storage medium | |
| CN116596851A (en) | Industrial flaw detection method based on knowledge distillation and anomaly simulation | |
| CN115909353A (en) | Image binarization processing method and device | |
| CN117765480B (en) | Method and system for early warning migration of wild animals along road | |
| CN118397367A (en) | Tampering detection method based on convolution vision Mamba | |
| CN114897850B (en) | Method and equipment for detecting surface defect of strip steel by using supervised Dropout | |
| CN115861306B (en) | Industrial product abnormality detection method based on self-supervision jigsaw module | |
| CN113807541B (en) | Fairness repair method, system, equipment and storage medium for decision system | |
| CN114862803A (en) | Industrial image anomaly detection method based on fast Fourier convolution | |
| CN111582446B (en) | System for neural network pruning and neural network pruning processing method | |
| CN113591856A (en) | Bill picture processing method and device | |
| CN116416486A (en) | Image recognition method and system | |
| CN113034353A (en) | Essential image decomposition method and system based on cross convolution neural network | |
| CN113222843A (en) | Image restoration method and related equipment thereof |
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 |