CN115240031B - Board surface defect generation method and system based on generation countermeasure network - Google Patents
Board surface defect generation method and system based on generation countermeasure network Download PDFInfo
- Publication number
- CN115240031B CN115240031B CN202210859013.9A CN202210859013A CN115240031B CN 115240031 B CN115240031 B CN 115240031B CN 202210859013 A CN202210859013 A CN 202210859013A CN 115240031 B CN115240031 B CN 115240031B
- Authority
- CN
- China
- Prior art keywords
- image
- defect
- defect detection
- background image
- training
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 230000007547 defect Effects 0.000 title claims abstract description 235
- 238000000034 method Methods 0.000 title claims abstract description 56
- 238000001514 detection method Methods 0.000 claims abstract description 137
- 238000012549 training Methods 0.000 claims abstract description 78
- 238000012795 verification Methods 0.000 claims abstract description 27
- 238000013507 mapping Methods 0.000 claims abstract description 23
- 230000002194 synthesizing effect Effects 0.000 claims abstract description 8
- 239000000463 material Substances 0.000 claims description 53
- 238000004458 analytical method Methods 0.000 claims description 8
- 230000001174 ascending effect Effects 0.000 claims description 6
- 230000015572 biosynthetic process Effects 0.000 claims description 6
- 238000003786 synthesis reaction Methods 0.000 claims description 6
- 230000002950 deficient Effects 0.000 claims description 4
- 230000001502 supplementing effect Effects 0.000 claims description 4
- 230000000694 effects Effects 0.000 abstract description 3
- 238000010801 machine learning Methods 0.000 abstract description 2
- 238000013434 data augmentation Methods 0.000 abstract 1
- 230000008485 antagonism Effects 0.000 description 9
- 238000005516 engineering process Methods 0.000 description 7
- 238000012545 processing Methods 0.000 description 5
- 238000010586 diagram Methods 0.000 description 3
- 238000007689 inspection Methods 0.000 description 3
- 238000004519 manufacturing process Methods 0.000 description 3
- 230000009286 beneficial effect Effects 0.000 description 2
- 238000004590 computer program Methods 0.000 description 2
- 239000013078 crystal Substances 0.000 description 2
- 230000003287 optical effect Effects 0.000 description 2
- 239000002023 wood Substances 0.000 description 2
- 238000004364 calculation method Methods 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 230000008713 feedback mechanism Effects 0.000 description 1
- 239000007888 film coating Substances 0.000 description 1
- 238000009501 film coating Methods 0.000 description 1
- 230000006870 function Effects 0.000 description 1
- 230000004927 fusion Effects 0.000 description 1
- 239000012535 impurity Substances 0.000 description 1
- 230000010365 information processing Effects 0.000 description 1
- 239000004579 marble Substances 0.000 description 1
- 238000005259 measurement Methods 0.000 description 1
- 238000004377 microelectronic Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- -1 offset Substances 0.000 description 1
- 238000004886 process control Methods 0.000 description 1
- 238000013441 quality evaluation Methods 0.000 description 1
- 239000004575 stone Substances 0.000 description 1
- 239000002344 surface layer Substances 0.000 description 1
- 238000012360 testing method Methods 0.000 description 1
Classifications
-
- 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
- G06V10/764—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
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Evolutionary Computation (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Computing Systems (AREA)
- Databases & Information Systems (AREA)
- Health & Medical Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Medical Informatics (AREA)
- Software Systems (AREA)
- Artificial Intelligence (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Multimedia (AREA)
- Image Analysis (AREA)
Abstract
The application relates to the technical field of machine learning, in particular to a method and a system for generating plate surface defects based on a generated countermeasure network, which aim at solving the problem that the prior art is difficult to obtain enough training samples and the detection precision of a defect detection model is poor. Acquiring an original plate image, and separating the original plate image into a background image set and a defect characteristic set; identifying pattern features of each sample in the background image set, classifying the background image set into at least one pattern field based on the pattern features; constructing a mapping relation between the pattern domain and the defect characteristic; synthesizing the background image set and the defect feature set into a training set and a verification set according to the mapping relation; the method and the device have the effects of data augmentation and are used for training the defect detection model and improving the defect detection precision.
Description
Technical Field
The application relates to the technical field of machine learning, in particular to a method and a system for generating plate surface defects based on a generated countermeasure network.
Background
Industrial automation refers to the collective term for information processing and process control of machine equipment or production processes to achieve measurement, manipulation, etc. according to an intended objective without direct human intervention. The automation technology is a method and technology for exploring and researching an automation process, and relates to a comprehensive technology in the technical fields of machinery, microelectronics, computers, machine vision and the like. Due to the need of the industrial revolution, automated technology has been vigorously developed. At the same time, automation technology has promoted industry progress, and automation technology has been widely used in the fields of machine manufacturing, electric power, construction, information technology and the like nowadays, and has become a main means for improving labor productivity.
The floor, i.e. the surface layer of the house floor or the floor, is made of timber or other materials, is an important component of the appearance of the house building, has great influence on the quality evaluation of the floor products and the purchase will of consumers in the floor industry, especially in the field of PVC floors, and therefore, the appearance quality inspection of the floor becomes an important ring in the production process of the floor.
At present, in the process of detecting the surface defects of the floor board, a floor image with the surface defects can be generally used as a training sample, an artificial intelligent model is trained, and finally the surface defects of the floor to be detected are detected through the trained model.
In carrying out the present application, the inventors have found that the above-described technique has at least the following problems:
before the model is actually put into use, a large number of floor images with surface defects need to be obtained as training samples, and under actual conditions, enough training samples are difficult to obtain, so that the problem of poor defect detection precision of the finally obtained model is easily caused.
Disclosure of Invention
In order to expand the sample source of the surface defect detection model and reduce the number of invalid samples at the same time, thereby improving the accuracy of the surface defect detection result, the application provides a plate surface defect generation method and system based on a generation countermeasure network.
In a first aspect, the present application provides a method for generating a surface defect of a board based on generation of an countermeasure network, which adopts the following technical scheme:
a method of generating sheet surface defects based on generating a countermeasure network, the method comprising the steps of:
acquiring an original plate image, and separating the original plate image into a background image set and a defect characteristic set;
identifying pattern features of each sample in the background image set, classifying the background image set into at least one pattern domain based on the pattern features;
performing material analysis on the background image based on the pattern domain, and constructing a mapping relation between the pattern domain and the defect feature based on a material analysis result;
synthesizing the background image set and the defect feature set into a defect detection sample set according to the mapping relation through a preset image generation model;
and training and acquiring a defect detection model based on the defect detection sample set.
By adopting the technical scheme, the background image and the defect feature on the original plate image are separated, the background image set and the defect feature set are respectively acquired, so that the defect feature originally fixed on a certain specific background can be synthesized on other background images, the number of samples finally used for training the defect detection model is enlarged, in addition, the background image set is classified based on the material features, the classified background image set and the defect feature set are mapped, the possibility that the defect feature which is not matched with the material of the background image is generated on the background image when the defect detection sample is synthesized is reduced, the effectiveness of the defect detection sample is improved, and the accuracy of the defect detection model in application is improved finally.
In a specific implementation manner, before the synthesizing the background image set and the defect feature set into the training set and the verification set according to the mapping relationship through a preset image generation model, the method further includes:
generating a training sample set and a verification sample set based on the pre-acquired original plate image;
and training a preset initial countermeasure network generation model through the training sample set and the verification sample set to acquire the image generation model.
By adopting the technical scheme, the training sample set and the verification sample set are generated through the original plate image, the training of the image generation model through the training sample set and the verification sample set is facilitated, the synthesis of different background images and style features in the original plate image is facilitated by means of the countermeasure network generation model, wherein the style features are defect features, the efficiency of image synthesis is improved, and the number of samples for training the defect detection model is also facilitated.
In a specific implementation manner, after the background image set and the defect feature set are synthesized into the training set and the verification set according to the mapping relationship through a preset image generation model, the method further includes:
generating a feedback training sample set and a feedback verification sample set based on the defect detection sample set;
training and updating the image generation model through the feedback training sample set and the feedback verification sample set;
replacing the initial image generation model with the updated image generation model.
By adopting the technical scheme, after the initial image generation model is acquired, the generated defect detection sample set is fed back to the initial image generation model as a training sample for training in the process of using the initial image model for image generation, so that the efficiency and the generation quality of the image generation model for image generation are further improved, and further the accuracy of defect detection is further improved.
In a specific embodiment, the method further comprises:
generating a random background image based on the material characteristics;
and supplementing the random background image to the background image set, wherein the duty ratio of the random background image in a material domain corresponding to the background image set does not exceed a preset random threshold value.
By adopting the technical scheme, the random background images are generated based on the material characteristics, so that the quantity of training samples is further enlarged through the images generated by artificial randomness, the occupation ratio of the random background images in the background image set is limited, the effectiveness of the training samples is controlled, and the possibility of poor precision of a final defect detection model caused by excessive random background images is avoided.
In a specific embodiment, the method further comprises:
in the training process of the defect detection model, acquiring a hierarchical defect detection model based on the number of defect detection samples used by the defect detection model;
and (3) arranging the grading defect detection models in an ascending order based on the number of the used defect detection samples, and connecting the grading defect detection models in series to form a defect detection system.
Through adopting above-mentioned technical scheme, in the in-process of training the defect detection model, the precision of training the model has been decided to the sample quantity that the training model operated, but a large amount of training samples also can lead to the training inefficiency of defect detection model, consequently acquire the defect detection model in grades, help according to the defect detection model of the suitable precision of demand acquisition, thereby improve the acquisition efficiency of defect detection model, establish ties the defect detection model of different precision as defect detection system, help utilizing defect detection system to carry out the progressively increasing detection of precision to panel image, thereby improve detection efficiency when guaranteeing the testing result precision.
In a specific embodiment, after the grading defect detection models are arranged in an ascending order based on the number of the defect detection samples used and are connected in series as a defect detection system, the method further includes:
acquiring a sheet material image to be detected, inputting the sheet material image to be detected into a defect detection system, and performing hierarchical detection according to the arrangement sequence of the defect detection system;
and when detecting that the surface defect exists in the image of the plate to be detected, stopping defect detection and marking the plate to be detected as a defective product.
Through adopting above-mentioned technical scheme, utilize defect detecting system to carry out the progressively increasing detection of precision to panel image, help improving detection efficiency when guaranteeing panel detection efficiency precision.
In a specific embodiment, after the original plate image is obtained and separated into the background image set and the defect feature set, the method further includes:
randomly adjusting image parameters of the background image set and the defect feature set;
the random adjustment includes a single image parameter adjustment and/or a combined adjustment of at least two image parameters.
By adopting the technical scheme, the image parameters in the background image set and the defect feature set are randomly adjusted, so that the number of samples is enlarged, and the adjustment comprises single random adjustment and combination of multiple random adjustments, so that the adjustment operation range is enlarged, and the number of samples is further enlarged.
In a second aspect, the application provides a sheet surface defect generating system based on generating an antagonism network, which adopts the following technical scheme:
a sheet surface defect generation system based on generating a countermeasure network, the system comprising:
the original image module is used for acquiring an original plate image and separating the original plate image into a background image set and a defect characteristic set;
a background classification module for identifying a texture feature of each sample in the background image set, classifying the background image set into at least one texture domain based on the texture feature;
the feature mapping module is used for constructing the mapping relation between the material domain and the defect feature;
the sample synthesis module is used for synthesizing the background image set and the defect feature set into a defect detection sample set according to the mapping relation through a preset image generation model;
and the detection model module is used for training and acquiring a defect detection model based on the defect detection sample set.
By adopting the technical scheme, the background image and the defect feature on the original plate image are separated, the background image set and the defect feature set are respectively acquired, so that the defect feature originally fixed on a certain specific background can be synthesized on other background images, the number of samples finally used for training the defect detection model is enlarged, in addition, the background image set is classified based on the material features, the classified background image set and the defect feature set are mapped, the possibility that the defect feature which is not matched with the material of the background image is generated on the background image when the defect detection sample is synthesized is reduced, the effectiveness of the defect detection sample is improved, and the accuracy of the defect detection model in application is improved finally.
In a third aspect, the present application provides an intelligent terminal, which adopts the following technical scheme:
a smart terminal comprising a processor and a memory having stored therein at least one instruction, at least one program, code set or instruction set loaded and executed by the processor to implement a method of generating sheet surface defects based on generating an countermeasure network as in any of the first aspects.
By adopting the technical scheme, the processor in the intelligent terminal can realize the method for generating the plate surface defects based on the generation countermeasure network according to the related computer program stored in the memory, so that the sample sources of the surface defect detection model are enlarged, and meanwhile, the number of invalid samples is reduced, so that the accuracy of the surface defect detection result is improved.
In a fourth aspect, the present application provides a computer readable storage medium, which adopts the following technical scheme:
a computer readable storage medium having stored therein at least one instruction, at least one program, code set, or instruction set loaded and executed by a processor to implement a method of generating sheet material surface defects based on generating an countermeasure network as in any of the first aspects.
By adopting the technical scheme, the corresponding program can be stored, so that the sample source of the surface defect detection model is enlarged, and meanwhile, the number of invalid samples is reduced, so that the accuracy of the surface defect detection result is improved.
In summary, the present application includes at least one of the following beneficial technical effects:
1. separating the background image and the defect feature on the original plate image is beneficial to respectively acquiring a background image set and a defect feature set, so that the defect feature originally fixed on a certain specific background can be synthesized on other background images, the number of samples finally used for training a defect detection model is enlarged, in addition, the background image set is classified based on material features, the classified background image set and the defect feature set are mapped, the possibility that defect features which are not matched with the material of the background image are generated on the background image when the defect detection sample is synthesized is reduced, the effectiveness of the defect detection sample is improved, and finally the accuracy of the defect detection model in application is improved;
2. generating a training sample set and a verification sample set through the original plate image, training an image generation model through the training sample set and the verification sample set, and combining different background images and style features in the original plate image by means of the countermeasure network generation model, wherein the style features are defect features, so that the image combination efficiency is improved, and the number of samples for training the defect detection model is increased;
3. in the process of training the defect detection model, the number of samples calculated by the training model determines the accuracy of the training model, but a large number of training samples also lead to low training efficiency of the defect detection model, so that the defect detection model is obtained in a grading manner, the defect detection model with proper accuracy is obtained according to requirements, the obtaining efficiency of the defect detection model is improved, the defect detection models with different accuracy are connected in series to form a defect detection system, the defect detection system is used for detecting the panel image with gradually increased accuracy, and the detection efficiency is improved while the accuracy of a detection result is ensured.
Drawings
In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings required for the description of the embodiments will be briefly described below, and it is apparent that the drawings in the following description are only some embodiments of the present application, and other drawings may be obtained according to these drawings without inventive effort for a person skilled in the art.
FIG. 1 is a method flow diagram of a sheet material surface defect generation method based on generation of an countermeasure network, shown in an embodiment of the application;
FIG. 2 is a system flow diagram of a sheet material surface defect generation system based on generation of an antagonism network, shown in an embodiment of the present application;
fig. 3 is a schematic structural diagram of an intelligent terminal according to an embodiment of the present application.
Detailed Description
The present embodiments are to be considered as merely illustrative and not restrictive, and modifications may occur to those skilled in the art upon reading the present specification and may be made without inventive faculty thereto, and it is to be understood that the embodiments are to be considered as part of, but not all of, the embodiments of the present application are to be considered as being explicitly described in connection with fig. 1-3 of the embodiments of the present application insofar as they are protected by the patent law within the scope of the appended claims.
The embodiment of the application provides a plate surface defect generation method based on a generation countermeasure network, which can be applied to an intelligent plate quality inspection process, wherein an execution subject can be an intelligent terminal and is assisted by a plate image acquisition device arranged on a quality inspection line.
The process flow shown in fig. 1 will be described in detail with reference to the specific embodiments, and the following may be included:
step 101, acquiring an original plate image, and separating the original plate image into a background image set and a defect characteristic set.
In implementation, the intelligent terminal acquires an original plate image firstly, wherein the original plate image can be a real plate image directly acquired by a technician through an image acquisition device, and the real plate image comprises a plate image with surface defects and a plate image without surface defects. After the original plate image is obtained, the intelligent terminal can pick out the defect image in the original plate image, mark the defect image by combining the defect information, and finally store the defect image as the defect characteristic. Therefore, the intelligent terminal can integrate the acquired defect characteristics into a defect characteristic set, and can store the plate image without defects and the plate image without surface defects as a background image together after the defect image is extracted, and integrate the plate image with the surface defects into the background image set.
At this time, the intelligent terminal may acquire a background image set and a defect image set of the board, so that the defect image is no longer fixed on a specific board.
In one embodiment, to expand the number of training samples, step 101 may further include the following processing steps: randomly adjusting image parameters of a background image set and a defect feature set; the random adjustment includes a single image parameter adjustment and/or a combined adjustment of at least two image parameters.
In particular, the stochastic processing may include one or more of brightness adjustment, contrast adjustment, flip adjustment, rotation adjustment, and gaussian noise perturbation.
Thus, the sample size can be increased based on the existing material by calculation.
Step 102, identifying pattern features of each sample in the background image set, classifying the background image set into at least one pattern field based on the pattern features.
In an implementation, after the intelligent terminal acquires the background image set, the pattern feature of each sample in the background image set may be extracted, and the background image set may be classified into a plurality of pattern fields based on the pattern feature. Specifically, the intelligent terminal may compare the texture features of the background image sample with preset pattern features by extracting the texture features of the background image sample, mark the background image sample with the texture features having the highest similarity to the texture features, and classify the whole background image set according to a preset texture domain after marking each background image sample, for example: the method comprises A, B, C three background image samples, wherein the corresponding material characteristics are marble plates, wood plates and plastic plates respectively, and A, B, C three background image samples can be respectively classified into three patterns of a stone background, a wood background and a plastic background.
In this way, it is possible to finely classify the background image samples in the background image set.
Step 103, carrying out material analysis on the background image set based on the pattern domain, and constructing a mapping relation between the pattern domain and the defect feature based on a material analysis result;
in implementation, after classifying the background image samples in the background image set, the intelligent terminal can establish the relationship between the defect features in the defect feature set and the material features corresponding to the material domain due to the fact that the occurrence of part of surface defects has a strong association relationship with the material, so as to remove the occurrence of the situation that the defect features are matched with the background images in the unmatched material domain. Specifically, for example, the defect feature set includes at least: the defects such as pits, offset, folds, scratches, broken sheets, crystal points, holes, bubbles, color escape, impurities, offset, film coating, dark bubbles and the like are usually generated on the plastic plate, so that the mapping relation between the crystal point defects and the plastic background can be established, and other conditions are similar and are not repeated.
Step 104, synthesizing a background image set and a defect feature set into a defect detection sample set according to a mapping relation through a preset image generation model;
in implementation, the intelligent terminal may perform image fusion on the background image sample in the background image set and the defect feature in the defect feature set by generating the countermeasure network image generation model based on the mapping relation described in step 103. Specifically, the intelligent terminal takes a background image sample as a background, and takes defect characteristics as style characteristics to be fused into the background image sample.
Thus, the intelligent terminal can finally obtain the defect detection sample set generated according to the specified mapping relation.
In one embodiment, to implement image composition, step 104 may be preceded by the following process: generating a training sample set and a verification sample set based on the pre-acquired original plate image; training a preset initial countermeasure network generation model through a training sample set and a verification sample set to acquire an image generation model.
Thus, the intelligent terminal can be trained in advance based on the original plate image to acquire the image generation model, so that the efficiency of image synthesis is improved.
In one embodiment, to optimize the generation quality of the image generation model, step 104 may further include the following processing: generating a feedback training sample set and a feedback verification sample set based on the defect detection sample set; training and updating the image generation model through a feedback training sample set and a feedback verification sample set; replacing the initial image generation model with the updated image generation model.
In this way, the intelligent terminal is helpful to continuously optimize the image generation model through a feedback mechanism so as to improve the computing capability of the image generation model.
In one embodiment, to expand the number of training samples, the method may further comprise the following processing steps: generating a random background image based on the material characteristics; and supplementing the random background image into a background image set, wherein the duty ratio of the random background image in a material domain corresponding to the background image set does not exceed a preset random threshold value.
Specifically, the random threshold may be set to 10%, 15%, 20%, or the like, and the specific value is set by a technician according to the actual situation.
In this way, it is helpful to further expand the number of training samples by means of a random background.
Step 105, training and obtaining a defect detection model based on the defect detection sample set.
In implementation, after the intelligent terminal acquires the defect detection sample set, the defect detection model can be trained by the defect detection sample set, and finally, the defect detection model can be updated.
In one embodiment, the method may further comprise the following processing steps: in the training process of the defect detection model, acquiring a hierarchical defect detection model based on the number of defect detection samples used by the defect detection model; the hierarchical defect detection models are arranged in ascending order based on the number of the used defect detection samples and are connected in series to form a defect detection system; acquiring a sheet material image to be detected, inputting the sheet material image to be detected into a defect detection system, and performing hierarchical detection according to the arrangement sequence of the defect detection system; and when detecting that the surface defect exists in the image of the plate to be detected, stopping defect detection and marking the plate to be detected as a defective product.
Thus, the training efficiency of the defect detection model is improved while the surface defect detection effect of the plate is ensured.
Based on the same technical concept, the embodiment of the application also provides a plate surface defect generating system based on generating an antagonism network, which comprises the following steps:
the original image module is used for acquiring an original plate image and separating the original plate image into a background image set and a defect characteristic set;
the background classification module is used for identifying the material characteristics of each sample in the background image set and classifying the background image set into at least one material domain based on the material characteristics;
the feature mapping module is used for constructing a mapping relation between the material domain and the defect feature;
the sample synthesis module is used for synthesizing the background image set and the defect feature set into a defect detection sample set according to a mapping relation through a preset image generation model;
and the detection model module is used for training and acquiring a defect detection model based on the defect detection sample set.
In one embodiment, a sheet material surface defect generation system based on generating an antagonism network further comprises:
the generation model training module is used for generating a training sample set and a verification sample set based on the pre-acquired original plate image; training a preset initial countermeasure network generation model through a training sample set and a verification sample set to acquire an image generation model.
In one embodiment, a sheet material surface defect generation system based on generating an antagonism network further comprises:
the feedback updating module is used for generating a feedback training sample set and a feedback verification sample set based on the defect detection sample set; training and updating the image generation model through a feedback training sample set and a feedback verification sample set; replacing the initial image generation model with the updated image generation model.
In one embodiment, a sheet material surface defect generation system based on generating an antagonism network further comprises:
the random background module is used for generating a random background image based on the material characteristics; and supplementing the random background image into a background image set, wherein the duty ratio of the random background image in a material domain corresponding to the background image set does not exceed a preset random threshold value.
In one embodiment, a sheet material surface defect generation system based on generating an antagonism network further comprises:
the hierarchical detection system module is used for acquiring a hierarchical defect detection model based on the number of defect detection samples used by the defect detection model in the training process of the defect detection model; the hierarchical defect detection models are arranged in ascending order based on the number of the used defect detection samples and are connected in series to form a defect detection system.
In one embodiment, a sheet material surface defect generation system based on generating an antagonism network further comprises:
the grading detection module is used for acquiring the image of the plate to be detected, inputting the image of the plate to be detected into the defect detection system, and carrying out grading detection according to the arrangement sequence of the defect detection system; and when detecting that the surface defect exists in the image of the plate to be detected, stopping defect detection and marking the plate to be detected as a defective product.
In one embodiment, a sheet material surface defect generation system based on generating an antagonism network further comprises:
the background adjusting module is used for randomly adjusting image parameters of the background image set and the defect characteristic set; the random adjustment includes a single image parameter adjustment and/or a combined adjustment of at least two image parameters.
The embodiment of the application also discloses an intelligent terminal, which comprises a memory and a processor, wherein the memory stores a computer program which can be loaded by the processor and execute the method for generating the sheet surface defects based on the generation countermeasure network.
Based on the same technical concept, the embodiment of the application also discloses a computer readable storage medium, which comprises the steps in the sheet material surface defect generation method flow based on the generation countermeasure network when being loaded and executed by a processor.
The computer-readable storage medium includes, for example: a U-disk, a removable hard disk, a Read-Only Memory (ROM), a random access Memory (Random Access Memory, RAM), a magnetic disk, or an optical disk, or other various media capable of storing program codes.
It will be clearly understood by those skilled in the art that, for convenience and simplicity of description, only the above-mentioned division of each functional module is illustrated, in practical application, the above-mentioned functional allocation may be performed by different functional modules according to needs, that is, the internal structure of the device is divided into different functional modules to perform all or part of the above-mentioned functions, and the specific working processes of the above-mentioned system, device and unit may refer to the corresponding processes in the foregoing method embodiments, which are not repeated herein.
The integrated units, if implemented in the form of software functional units and sold or used as stand-alone products, may be stored in a computer readable storage medium. Based on such an understanding that the technical solution of the present application is essentially or partly contributing to the prior art or all or part of the technical solution may be embodied in the form of a software product stored in a storage medium, comprising several instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) or a processor (processor) to perform all or part of the steps of the method of the embodiments of the present application. And the aforementioned storage medium includes: various media capable of storing program codes, such as a U disk, a mobile hard disk, a read-only memory, a random access memory, a magnetic disk or an optical disk.
The foregoing embodiments are only used to describe the technical scheme of the present application in detail, but the descriptions of the foregoing embodiments are only used to help understand the method and the core idea of the present application, and should not be construed as limiting the present application. Variations or alternatives, which are easily conceivable by those skilled in the art, are included in the scope of the present application.
Claims (9)
1. A plate surface defect generation method based on a generation countermeasure network is characterized by comprising the following steps of: the method comprises the following steps:
acquiring an original plate image, and separating the original plate image into a background image set and a defect characteristic set;
identifying pattern features of each sample in the background image set, classifying the background image set into at least one pattern domain based on the pattern features;
performing material analysis on the background image set based on the pattern domain, and constructing a mapping relation between the pattern domain and the defect feature based on a material analysis result;
generating a training sample set and a verification sample set based on the pre-acquired original plate image;
training a preset initial countermeasure network generation model through the training sample set and the verification sample set to acquire the image generation model;
synthesizing the background image set and the defect feature set into a training set and a verification set according to the mapping relation through the image generation model;
and training and acquiring a defect detection model based on the defect detection sample set.
2. A sheet material surface defect generation method based on generation of an countermeasure network according to claim 1, wherein: after the background image set and the defect feature set are synthesized into the training set and the verification set according to the mapping relation through a preset image generation model, the method further comprises the following steps:
generating a feedback training sample set and a feedback verification sample set based on the defect detection sample set;
training and updating the image generation model through the feedback training sample set and the feedback verification sample set;
replacing the initial image generation model with the updated image generation model.
3. A method of generating sheet surface defects based on generation of a countermeasure network as recited in claim 1, further comprising:
generating a random background image based on the pattern features;
and supplementing the random background image to the background image set, wherein the duty ratio of the random background image in the pattern domain corresponding to the background image set does not exceed a preset random threshold value.
4. A sheet material surface defect generation method based on generation of an countermeasure network according to claim 1, wherein: the method further comprises the steps of:
in the training process of the defect detection model, acquiring a hierarchical defect detection model based on the number of defect detection samples used by the defect detection model;
and (3) arranging the grading defect detection models in an ascending order based on the number of the used defect detection samples, and connecting the grading defect detection models in series to form a defect detection system.
5. A method of generating sheet surface defects based on generation of a countermeasure network as recited in claim 4, wherein: the step of arranging the classified defect detection models in an ascending order based on the number of the used defect detection samples and connecting the classified defect detection models in series to form a defect detection system further comprises:
acquiring a sheet material image to be detected, inputting the sheet material image to be detected into a defect detection system, and performing hierarchical detection according to the arrangement sequence of the defect detection system;
and when detecting that the surface defect exists in the image of the plate to be detected, stopping defect detection and marking the plate to be detected as a defective product.
6. A sheet material surface defect generation method based on generation of an countermeasure network according to claim 1, wherein: the step of obtaining the original plate image, after separating the original plate image into a background image set and a defect feature set, further comprises:
randomly adjusting image parameters of the background image set and the defect feature set;
the random adjustment includes a single image parameter adjustment and/or a combined adjustment of at least two image parameters.
7. A sheet surface defect generation system based on generation of a countermeasure network, characterized in that: the system comprises:
the original image module is used for acquiring an original plate image and separating the original plate image into a background image set and a defect characteristic set;
a background classification module for identifying pattern features of each sample in the background image set, classifying the background image set into at least one pattern field based on the pattern features;
the feature mapping module is used for carrying out material analysis on the background image based on the pattern domain and constructing a mapping relation between the pattern domain and the defect feature based on a material analysis result;
the sample synthesis module is used for synthesizing the background image set and the defect feature set into a defect detection sample set according to the mapping relation through a preset image generation model;
the detection model module is used for training and acquiring a defect detection model based on the defect detection sample set;
the generation model training module is used for generating a training sample set and a verification sample set based on the pre-acquired original plate image; training a preset initial countermeasure network generation model through a training sample set and a verification sample set to acquire an image generation model.
8. A smart terminal comprising a processor and a memory, wherein the memory stores at least one instruction, at least one program, a set of codes, or a set of instructions, the at least one instruction, the at least one program, the set of codes, or the set of instructions being loaded and executed by the processor to implement a method of generating sheet surface defects based on generating an countermeasure network as claimed in any one of claims 1 to 7.
9. A computer-readable storage medium having stored therein at least one instruction, at least one program, code set, or instruction set, the at least one instruction, the at least one program, the code set, or instruction set being loaded and executed by a processor to implement a method of generating sheet material surface defects based on generating an countermeasure network as claimed in any one of claims 1 to 6.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202210859013.9A CN115240031B (en) | 2022-07-21 | 2022-07-21 | Board surface defect generation method and system based on generation countermeasure network |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202210859013.9A CN115240031B (en) | 2022-07-21 | 2022-07-21 | Board surface defect generation method and system based on generation countermeasure network |
Publications (2)
Publication Number | Publication Date |
---|---|
CN115240031A CN115240031A (en) | 2022-10-25 |
CN115240031B true CN115240031B (en) | 2023-10-27 |
Family
ID=83674186
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202210859013.9A Active CN115240031B (en) | 2022-07-21 | 2022-07-21 | Board surface defect generation method and system based on generation countermeasure network |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN115240031B (en) |
Families Citing this family (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN115937595A (en) * | 2022-12-20 | 2023-04-07 | 中交公路长大桥建设国家工程研究中心有限公司 | Bridge apparent anomaly identification method and system based on intelligent data processing |
Citations (9)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN108447054A (en) * | 2018-03-22 | 2018-08-24 | 北京木业邦科技有限公司 | Defects in timber sample acquiring method, device, electronic equipment and storage medium |
CN109829483A (en) * | 2019-01-07 | 2019-05-31 | 鲁班嫡系机器人(深圳)有限公司 | Defect recognition model training method, device, computer equipment and storage medium |
CN110570316A (en) * | 2018-08-31 | 2019-12-13 | 阿里巴巴集团控股有限公司 | method and device for training damage recognition model |
CN111311544A (en) * | 2020-01-19 | 2020-06-19 | 无锡赛默斐视科技有限公司 | Floor defect detection method based on deep learning |
CN112017182A (en) * | 2020-10-22 | 2020-12-01 | 北京中鼎高科自动化技术有限公司 | Industrial-grade intelligent surface defect detection method |
CN113205176A (en) * | 2021-04-19 | 2021-08-03 | 重庆创通联达智能技术有限公司 | Method, device and equipment for training defect classification detection model and storage medium |
CN113283541A (en) * | 2021-06-15 | 2021-08-20 | 无锡锤头鲨智能科技有限公司 | Automatic floor sorting method |
CN114092386A (en) * | 2020-05-26 | 2022-02-25 | 富士通株式会社 | Defect detection method and apparatus |
CN114387230A (en) * | 2021-12-28 | 2022-04-22 | 北京科技大学 | PCB defect detection method based on re-verification detection |
-
2022
- 2022-07-21 CN CN202210859013.9A patent/CN115240031B/en active Active
Patent Citations (9)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN108447054A (en) * | 2018-03-22 | 2018-08-24 | 北京木业邦科技有限公司 | Defects in timber sample acquiring method, device, electronic equipment and storage medium |
CN110570316A (en) * | 2018-08-31 | 2019-12-13 | 阿里巴巴集团控股有限公司 | method and device for training damage recognition model |
CN109829483A (en) * | 2019-01-07 | 2019-05-31 | 鲁班嫡系机器人(深圳)有限公司 | Defect recognition model training method, device, computer equipment and storage medium |
CN111311544A (en) * | 2020-01-19 | 2020-06-19 | 无锡赛默斐视科技有限公司 | Floor defect detection method based on deep learning |
CN114092386A (en) * | 2020-05-26 | 2022-02-25 | 富士通株式会社 | Defect detection method and apparatus |
CN112017182A (en) * | 2020-10-22 | 2020-12-01 | 北京中鼎高科自动化技术有限公司 | Industrial-grade intelligent surface defect detection method |
CN113205176A (en) * | 2021-04-19 | 2021-08-03 | 重庆创通联达智能技术有限公司 | Method, device and equipment for training defect classification detection model and storage medium |
CN113283541A (en) * | 2021-06-15 | 2021-08-20 | 无锡锤头鲨智能科技有限公司 | Automatic floor sorting method |
CN114387230A (en) * | 2021-12-28 | 2022-04-22 | 北京科技大学 | PCB defect detection method based on re-verification detection |
Non-Patent Citations (1)
Title |
---|
工业CT/DR检测系统在盆式绝缘子检测中的应用研究;尹奎龙等;山东电力技术(08);全文 * |
Also Published As
Publication number | Publication date |
---|---|
CN115240031A (en) | 2022-10-25 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN106952250B (en) | Metal plate strip surface defect detection method and device based on fast R-CNN network | |
WO2019117065A1 (en) | Data generation device, data generation method and data generation program | |
CN106934800B (en) | Metal plate strip surface defect detection method and device based on YOLO9000 network | |
CN109118482A (en) | A kind of panel defect analysis method, device and storage medium | |
CN116188475B (en) | Intelligent control method, system and medium for automatic optical detection of appearance defects | |
CN103439348B (en) | Remote controller key defect detection method based on difference image method | |
Mahalingam et al. | Pcb-metal: A pcb image dataset for advanced computer vision machine learning component analysis | |
CN104463199A (en) | Rock fragment size classification method based on multiple features and segmentation recorrection | |
CN115240031B (en) | Board surface defect generation method and system based on generation countermeasure network | |
CN106846313A (en) | Surface Flaw Detection method and apparatus | |
CN114862845B (en) | Defect detection method, device and equipment for mobile phone touch screen and storage medium | |
CN113222913B (en) | Circuit board defect detection positioning method, device and storage medium | |
CN207636508U (en) | A kind of defect of veneer detecting system and equipment | |
CN112070747A (en) | LED lamp bead defect detection method and device | |
WO2024002187A1 (en) | Defect detection method, defect detection device, and storage medium | |
CN113935988A (en) | Box body surface defect detection and identification method and device based on machine vision | |
TW201606854A (en) | Automatic calibration sample selection for die-to-database photomask inspection | |
CN108375353A (en) | A kind of log detection method, apparatus and equipment | |
CN114216911A (en) | Powder laying quality monitoring and controlling method in metal selective laser melting forming | |
CN106323985B (en) | Solid wood board quality detection method combining computer vision with self-learning behavior | |
CN113724223B (en) | YOLOv3 data set production method and system based on optical microscope | |
CN113822836B (en) | Method for marking an image | |
Ruediger-Flore et al. | CAD-based data augmentation and transfer learning empowers part classification in manufacturing | |
CN104504429B (en) | two-dimensional code generation method and device | |
Sachdeva et al. | Research Trends in Image Processing and Defect Detections |
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 |