CN115661071A - Composite material processing surface defect detection and evaluation method based on deep learning - Google Patents
Composite material processing surface defect detection and evaluation method based on deep learning Download PDFInfo
- Publication number
- CN115661071A CN115661071A CN202211311389.2A CN202211311389A CN115661071A CN 115661071 A CN115661071 A CN 115661071A CN 202211311389 A CN202211311389 A CN 202211311389A CN 115661071 A CN115661071 A CN 115661071A
- Authority
- CN
- China
- Prior art keywords
- defect
- evaluation
- training
- composite material
- model
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Images
Classifications
-
- 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
- Image Analysis (AREA)
Abstract
The invention belongs to the technical field related to precision machining detection, and discloses a composite material machining surface defect detection and evaluation method based on deep learning, which comprises the following steps: (1) Shooting a defect image, and carrying out data marking on the image; (2) Dividing a training set and a test set, and performing data enhancement processing on a data set; (3) inputting the training set into a deep learning model for training; (4) Inputting the test set into the trained model to obtain the category, position, area of the region and model evaluation index of the defect in the image; and (5) forming a mapping of the evaluation index to the evaluation score. The invention also discloses a corresponding system. The invention can realize high-accuracy and high-efficiency particle-reinforced composite material processing surface defect detection, realize quantitative evaluation of the detected defects, and effectively guide the selection of optimal process parameters, thereby being particularly suitable for application occasions of particle-reinforced composite material processing surface defect detection and evaluation.
Description
Technical Field
The invention belongs to the technical field related to precision machining detection, and particularly relates to a composite material machining surface defect detection and evaluation method based on deep learning.
Background
Along with the application of spacecrafts in space and near space, the structural volume, weight and manufacturing precision of core optical elements of various systems are more and more emphasized, and the lightening and ultra-precise manufacturing of high-performance optical elements of spacecrafts are the necessary way to ensure the high-quality comprehensive performance of spaceflight equipment. A single material is not always able to meet certain requirements in the industrial field, so particle-reinforced composites that are able to meet these requirements are widely used.
The composite material has small thermal expansion coefficient, high strength, high rigidity, excellent wear resistance, corrosion resistance, high temperature resistance and other excellent performances, and is an ideal novel light material. The composite material has the defects of easy formation of various forms in the processing process due to large physical and mechanical property difference between the metal matrix and the reinforced phase, and the material properties are seriously influenced by furrows, cracks, particle breakage, bulges, interface stripping and the like.
In the prior art, methods for detecting defects on the processed surface of a particle-reinforced material mainly comprise traditional manual visual detection, ultrasonic detection, X-ray detection, high-frequency pulse eddy current detection and the like. Wherein, the manual detection is biased to subjective judgment, and simultaneously, the detection of minor defects has low efficiency and accuracy; ultrasonic detection is a detection technology which is common and widely used for the composite material at present, but the defect display is not intuitive enough, so that the qualitative and quantitative determination of the defect is difficult, a coupling agent is needed, and the method is mainly suitable for internal defect detection; eddy current testing requires that the material itself be conductive and require analysis by a practitioner. Most of the methods need manual or semi-manual defect judgment, resulting in low efficiency.
Accordingly, there is a need in the art for further research and improvement to better meet the requirements of high-precision and high-efficiency detection and quantitative evaluation of the machined surface of the composite material.
Disclosure of Invention
In view of the above defects or requirements of the prior art, an object of the present invention is to provide a method for detecting and evaluating a defect on a machined surface of a composite material based on deep learning, wherein the relevant characteristics of the defect on the machined surface of the composite material are fully considered, a deep learning algorithm is selected, and operations such as training, testing and evaluation are designed in a targeted manner, so that compared with the prior art, the detection capability of a model on the defect on the machined surface of the composite material can be further improved, and a high-accuracy and high-efficiency detection of the defect on the machined surface and a more comprehensive and quantitative evaluation result can be obtained.
In order to achieve the above object, according to one aspect of the present invention, there is provided a method for detecting and evaluating a defect on a composite material processing surface based on deep learning, the method comprising:
step one, image acquisition and labeling
Shooting images of the processed surface of the composite material, summarizing the obtained images into an image set and executing defect labeling, thereby forming an image data set;
step two, image data set division and data enhancement
Dividing the image data set formed in the step one into a training set and a testing set, respectively training the model and the testing model, and simultaneously carrying out image data enhancement processing on the divided training set to enlarge the scale of the training set;
step three, training a defect detection model
Inputting the training set obtained in the step two into a deep learning model for model training;
step four, testing a defect detection model
Inputting the test set obtained in the second step into a trained deep learning model for testing to obtain information such as defect category, defect position, defect depth, defect area ratio, length and width of a minimum circumscribed rectangle of a defect area and the like in the image, thereby obtaining a corresponding defect evaluation index value;
step five, defect evaluation
And forming a mapping relation between the evaluation indexes and the evaluation scores based on the defect evaluation index values obtained in the fourth step and by combining a preset defect evaluation score criterion, thereby completing the whole detection and evaluation process.
As a further preference, in step one, the number of images obtained is preferably not less than 800, and the corresponding number of each defect is preferably not less than 200; the defect labeling preferably comprises: defect type, coordinate information of defect region frame, defect instance boundary point, defect instance area, etc.
As a further preference, in step two, the division ratio of the training set and the test set is preferably 8:2, and rotation, scaling, cutting, mosaic, cutMix and other methods can be adopted to complete the image data enhancement processing of the training set.
As a further preference, for the deep learning model, it is preferable to set as follows: the backbone of the system uses a ResNet network to extract features and combines with an FPN network to output feature maps with different sizes; the propulses are generated through the RPN, and the Fast-RCNN carries out category prediction and position fine adjustment on the propulses generated by the RPN; and generating all types of masks by Mask branches, and extracting masks corresponding to prediction types.
As a further preference, in step three, the process of model training is preferably designed as follows:
performing K-means clustering on the defect marking frames of the training set to obtain a proper anchor size; extracting feature maps of different levels through a backbone network and a FPN network, obtaining the explosals through the RPN network, and then mapping the obtained explosals to the corresponding level feature maps to obtain the explosals feature maps, wherein the corresponding relation is as follows:
wherein k is 0 Is w.h = S 2 The mapped layer number, w and h are the width and the height of the propofol respectively;
in addition, the propusals feature graphs of different levels are converted into the same size through RoIAlign, class prediction and propusal offset prediction of the feature graphs are achieved through two full-connection layers and finally through two parallel full-connection layers, and input targets of Mask branches are propusals provided by RPNs during training.
Further preferably, in step three, the loss of model training preferably comprises RPN network loss, fast-RCNN loss, mask loss, wherein
The associated RPN loss function is designed to:
wherein, N cls Selecting the number of candidate frames for calculating loss for a picture, p i The probability of being a positive sample is predicted for the ith anchor,
1 in the case of positive samples and 0 in the case of negative samples reg Number of anchors points, t i To predict the ith anchor corresponding regressionThe parameters are set to be in a predetermined range,
the regression parameter of the GTBox corresponding to the ith anchor is taken as the regression parameter of the GTBox;
the relevant Fast-RCNN loss function is designed as:
L(p，u，t u ，v)＝L cls (p，u)+λ[u≥1]L loc (t u ，v)
wherein, t u V is a bounding box regression parameter corresponding to the real target for predicting the regression parameter corresponding to the category u;
the relevant Mask loss function is designed as:
L(m，n)＝L BCE (m，n)
wherein m is Mask corresponding to the prediction category, and n is GT Mask.
As a further preference, in step four, the process of the model test is preferably designed as follows:
obtaining a plurality of feature maps through a backbone and an FPN, generating corresponding explosals of each feature map through an RPN, and mapping the explosals onto the corresponding feature maps to obtain the explosals feature maps; and then, obtaining a prediction category and a related offset corresponding to a proposal through RoIAlign, two full connection layers and two parallel full connection layers, mapping the offset proposals output by the Fast-RCNN network back to a feature map, changing the size through RoIAlign, inputting the result to a Mask branch, and selecting a Mask corresponding to the target prediction category to map back to the original image.
As a further preference, in the fourth step, the defect area, the minimum bounding rectangle of the defect region, and the defect type can be obtained by the target detection algorithm based on deep learning, wherein the defect area can be obtained by scaling the pixel area and the scale in the image, and the minimum bounding rectangle of the defect region can be obtained by the rotation card shell algorithm.
As a further preference, in the fifth step, the defect evaluation score criterion may preferably be selected from mechanical properties, physical properties, chemical properties, service life and the like, wherein the mechanical properties further include yield strength, shear modulus, section shrinkage and the like, the physical properties further include electrical resistivity, thermal conductivity, refractive index and the like, the chemical properties further include corrosion resistance, oxidation resistance and the like, and these parameters are proportionally adjusted by weighting coefficients.
As a further preference, in step five, a mapping relationship between the evaluation index and the evaluation score is preferably established through an artificial neural network, wherein the defect category is embodied by weighting corresponding parameters of different types of defects.
According to another aspect of the present invention, there is also provided a corresponding deep learning-based composite material processing surface defect detection and evaluation method, wherein the system comprises:
an image acquisition and labeling module for capturing images of the composite material machined surface, summarizing the acquired images into an image set and performing defect labeling, thereby forming an image data set;
the image data set dividing and data enhancing module is used for dividing the formed image data set into a training set and a testing set, respectively training a model and the testing model, and simultaneously performing image data enhancing processing on the divided training set to enlarge the scale of the training set;
the defect detection model training module is used for inputting the obtained training set into the deep learning model for model training;
the defect detection model testing module is used for inputting the obtained testing set into a trained deep learning model for testing to obtain information such as defect types, defect positions, defect depths, defect areas, defect area ratios, length and width of a minimum circumscribed rectangle of a defect area and the like in an image so as to obtain corresponding defect evaluation index values;
and the defect evaluation module is used for forming a mapping relation between the evaluation index and the evaluation score based on the acquired defect evaluation index value and in combination with a preset defect evaluation score criterion, so that the whole detection and evaluation process is completed.
In general, the above technical solution conceived by the present invention has the following advantages compared to the prior art
Has the advantages that:
(1) Compared with the traditional manual visual detection, ultrasonic detection and X-ray detection, the defect identification efficiency and accuracy can be effectively improved, and the subjective interference caused by manual work or semi-manual work in the traditional method can be avoided;
(2) According to the method, the generalization capability of the defect detection model can be effectively improved through a large number of defect samples and a data enhancement mode;
(3) The method can effectively improve the recognition capability of the defects with different sizes and reduce the missing rate of the defects by combining the feature extraction network and the FPN network;
(4) The invention further provides a defect evaluation system which can make up the problem that the judgment standard of the defects is self-subjective judgment of detection personnel and a set of good quantitative evaluation method is not available, and the method is used for quantifying the defects to guide processing and can effectively promote the processing of the granular composite material to be developed towards intelligent manufacturing.
Drawings
FIG. 1 is an overall flow chart of a method for detecting and evaluating composite material processing surface defects based on deep learning according to the present invention;
FIG. 2 is a schematic diagram illustrating an exemplary defect detection model in accordance with a preferred embodiment of the present invention;
FIG. 3 is a schematic diagram of a feature extraction network for exemplary display incorporating FPNs in accordance with a preferred embodiment of the present invention;
FIG. 4 is a schematic flow diagram for an exemplary display defect evaluation phase in accordance with a preferred embodiment of the present invention;
FIG. 5 is a diagram illustrating actual outputs of a defect detection model in accordance with a preferred 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.
FIG. 1 is an overall flow chart of a composite material processing surface defect detection and evaluation method based on deep learning according to the invention. The invention will be explained in more detail below with reference to fig. 1.
First, the image acquisition and annotation step.
In this step, an image of the composite material work surface is taken, the obtained images are grouped into an image set and defect labeling is performed, thereby forming an image data set.
More specifically, an image of the surface of the particle-reinforced composite material may be taken, for example, by electron microscopy, the images obtained are combined into an image set, and the image set is defect labeled using suitable data labeling software to form an image data set.
For example, 1000 pictures of aluminum-based silicon carbide defects can be obtained by SEM, with the corresponding number of each defect being not less than 200. And (3) making a COCO data set through labelme, and labeling the defect type, the defect region frame information (left upper corner coordinate, height and width), the defect instance polygon boundary point and the defect area.
Next, image dataset partitioning and data enhancement steps are performed.
In this step, the previously formed image data set is divided into a training set and a test set, which are respectively used for training a model and a test model, and image data enhancement processing is performed on the divided training set to enlarge the scale of the training set.
More specifically, it is preferable that the picture data can be enhanced to 3000 pieces by rotating, scaling, flipping, cutMix, etc., and the picture is processed in 8: and 2, dividing the training set and the test set in proportion, namely 2400 training set pictures and 600 test set pictures.
Next, a defect detection model training step is performed.
In this step, the obtained training set is input to the deep learning model, and model training is performed.
More specifically, the present invention can use, for example, the Mask-RCNN algorithm, as shown in FIG. 2, and the model in this embodiment includes a feature extraction network (ResNet 50+ FPN), an RPN network, a Fast-RCNN network, and a Mask branch. As shown in fig. 3, defects of different sizes can be effectively detected using a feature extraction network comprising FPNs. In the training process, K-means clustering needs to be carried out on a training set defect labeling frame to obtain a proper anchor size, a backbone is pre-trained before the whole model is trained, and the problems of small data volume and slow network convergence are solved through transfer learning. In the training process, the output of the RPN is used by the Fast-RCNN and the Mask branch, and the proposals is mapped back to the corresponding characteristic diagram according to the size, and the mapping formula is as follows:
wherein k is 0 Is w · h = S 2 The mapped layer numbers w and h are the width and height of the propofol respectively.
Two different roialligns are used for size conversion of the prosassals feature map by the two branches, and the related weights need to be trained respectively.
According to a preferred embodiment of the present invention, the training total loss comprises RPN network loss, fast-RCNN loss and Mask loss.
RPN loss function:
wherein N is cls Selecting the number of candidate frames for calculating loss, p, for a picture i The probability of being a positive sample is predicted for the ith anchor,
1 when positive, negative0,N in the case of the sample reg Number of anchors points, t i To predict the regression parameters corresponding to the ith anchor,
the ith anchor corresponds to the regression parameters of the GTBox.
Fast-RCNN loss function:
L(p，u，t u ，v)＝L clS (p，u)+λ[u≥1]L loc (t u ，v)
wherein, t u V corresponds to the bounding box regression parameters of the real target for predicting the regression parameters corresponding to the category u.
Mask loss function:
L(m，n)＝L BCE (m，n)
wherein m is Mask corresponding to the prediction category, and n is GT Mask.
In addition, the model training environment can use a pytorch framework, hardware uses GeForce RTX 3080, gradient updating is carried out by adopting an SGD algorithm with momentum, momentum parameters are set to be 0.9, the learning rate is 0.0004, weight attenuation is 0.0001, 26 epochs are iterated totally, and gradient 0.1-ratio learning rate attenuation is carried out on 16 th and 22 th epochs.
Next, a defect detection model test procedure is performed.
In this step, the previously obtained test set is input into a trained deep learning model for testing, and information such as defect type, defect position, defect depth, defect area ratio, length and width of a minimum circumscribed rectangle of a defect area in an image is obtained, so that a corresponding defect evaluation index value is obtained.
More specifically, the weights obtained by training can be used to input a test set picture into a defect detection model, the test flow is as shown in fig. 3, a final prediction frame and a prediction category are obtained through a Fast-RCNN network, then the final prediction frame and the prediction category are used as input of Mask branches to obtain a corresponding Mask of the category, and then the corresponding Mask is mapped back to an original picture to obtain parameters such as a defect category, a position, an area and the like which are finally required, mAP is selected as a model evaluation index of target detection and image segmentation, the mAP is an average precision mean value of all categories, the larger the mAP value is, the better the performance of the model is represented, and if the mAP value of the final test set meets the requirements, the trained model can be used for detecting actual defect pictures.
Finally, there is a defect evaluation step.
In this step, a mapping relationship between the evaluation index and the evaluation score is formed based on the acquired defect evaluation index value and in combination with a preset defect evaluation score criterion, thereby completing the whole detection and evaluation process.
More specifically, the defect evaluation in the invention comprises selecting a defect evaluation index and an evaluation score criterion, and forming a mapping of the evaluation index to the evaluation score. The whole defect evaluation process is shown in fig. 4, wherein the defect evaluation indexes include defect area, defect area ratio, length and width of the minimum circumscribed rectangle of the defect area, defect type, defect depth, and the like. The defect types are directly obtained by a deep learning model; the indexes of the defect area, the defect area ratio, the length and the width of the minimum circumscribed rectangle of the defect area and the like are obtained by the output processing of a post-processing algorithm on the deep learning model; the defect depth and other indexes which cannot be directly and indirectly obtained from the two-dimensional image need to be obtained through other measuring ways. For example, the defect area can be obtained by converting the pixel area in the image and a scale bar, the minimum bounding rectangle of the defect area can be obtained by a rotating card shell algorithm, and the defect depth can be obtained by SEM measurement.
The evaluation score evaluation criterion can be selected from mechanical properties (yield strength, shear modulus, section shrinkage rate and the like), physical properties (resistivity, thermal conductivity, refractive index and the like), chemical properties (corrosion resistance, oxidation resistance and the like), service life and the like, and parameters can be selected according to the use background of the material. When a plurality of parameters are selected, normalization processing is required in consideration of different unit and numerical magnitude of different parameters, and proportion adjustment of the plurality of parameters is required through weighting coefficients. For example, when an aluminum-based silicon carbide material is processed to manufacture a space reflector, performance parameters such as reflection performance, thermal conductivity and yield strength meet the use requirements, the selected parameters are normalized to a [0,1] interval, the sum of weighting coefficients is 1, and a fraction formula can be expressed as:
score＝w 1 a+w 2 b+w 3 c
wherein a, b and c are values after normalization treatment of reflection performance, thermal conductivity and yield strength, and w 1 、w 2 、w 3 Is a weighting coefficient of the corresponding parameter, and w 1 +w 2 +w 3 ＝1。
According to another preferred embodiment of the present invention, the mapping relationship between the evaluation index and the evaluation score is established by an artificial neural network, i.e., the mapping weight between neurons and nonlinear activation functions is learned by a large amount of data. The defect types are embodied by weighting corresponding parameters of different types of defects, such as common particle breakage and scratch defects of aluminum-based silicon carbide, when neural network mapping is carried out, relevant parameters of the two defects are independently used as input neurons, and the weight representing the defect types is self-learned through back propagation.
In conclusion, the invention can realize high-accuracy and high-efficiency particle-reinforced composite material processing surface defect detection, realize quantitative evaluation of the detected defects and effectively guide the selection of optimal process parameters, thereby being particularly suitable for application occasions of particle-reinforced composite material surface defect detection and evaluation. .
It will be understood by those skilled in the art that the foregoing is only a preferred embodiment of the present invention, and is not intended to limit the invention, and that any modification, equivalent replacement, or improvement made within the spirit and principle of the present invention should be included in the scope of the present invention.
Claims (10)
1. A composite material processing surface defect detection and evaluation method based on deep learning is characterized by comprising the following steps:
step one, image acquisition and labeling
Shooting images of the processed surface of the composite material, summarizing the obtained images into an image set and executing defect labeling, thereby forming an image data set;
step two, dividing the image data set and enhancing the data
Dividing the image data set formed in the step one into a training set and a testing set, respectively training the model and the testing model, and simultaneously carrying out image data enhancement processing on the divided training set to enlarge the scale of the training set;
step three, training a defect detection model
Inputting the training set obtained in the step two into a deep learning model for model training;
step four, testing a defect detection model
Inputting the test set obtained in the second step into a trained deep learning model for testing to obtain information such as defect category, defect position, defect depth, defect area ratio, length and width of a minimum circumscribed rectangle of a defect area and the like in the image, thereby obtaining a corresponding defect evaluation index value;
step five, defect evaluation
And forming a mapping relation between the evaluation indexes and the evaluation scores based on the defect evaluation index values obtained in the step four and by combining a preset defect evaluation score criterion, thereby completing the whole detection and evaluation process.
2. The method for detecting and evaluating defects on composite material processing surfaces according to claim 1, wherein in step one, the number of images obtained is preferably not less than 800, and the corresponding number of each defect is preferably not less than 200; the defect labeling preferably comprises: defect type, coordinate information of defect region box, defect instance boundary point, defect instance area, etc.
3. The method for detecting and evaluating defects on composite material processing surfaces according to claim 1 or 2, wherein in step two, the training set and the test set are preferably divided into 8:2, and rotation, scaling, cutting, mosaic, cutMix and other methods can be adopted to complete the image data enhancement processing of the training set.
4. The method for detecting and evaluating defects on a composite material processing surface according to any one of claims 1 to 3, wherein the deep learning model is preferably set as follows: the backbone of the method uses a ResNet network to extract features and combines with an FPN network to output feature maps with different sizes; generating propulses through an RPN (resilient packet network), and performing category prediction and position fine adjustment on the propulses generated by the RPN through a Fast-RCNN (Fast-RCNN) network; and generating masks of all categories by Mask branches, and extracting masks corresponding to the prediction categories.
5. The method for detecting and evaluating the defects on the processed surface of the composite material according to any one of the claims 1 to 4, wherein in the third step, the model training process is preferably designed as follows:
performing K-means clustering on the defect marking frame of the training set to obtain a proper anchor size; extracting feature graphs of different levels through a backbone network and an FPN network, obtaining explosals through an RPN network, and mapping the obtained explosals feature graphs to corresponding level feature graphs to obtain the explosals feature graphs, wherein the corresponding relation is as follows:
wherein k is 0 Is w · h = S 2 The mapped layer number, w and h are the width and the height of the propofol respectively;
in addition, the propusals feature graphs of different levels are converted into the same size through RoIAlign, class prediction and propusal offset prediction of the feature graphs are achieved through two full-connection layers and finally through two parallel full-connection layers, and input targets of Mask branches are propusals provided by RPNs during training.
6. The method according to claim 5, wherein the loss of model training in step three preferably comprises RPN network loss, fast-RCNN loss and Mask loss, wherein
The associated RPN loss function is designed to:
wherein N is cls Selecting the number of candidate frames for calculating loss, p, for a picture i The probability of being a positive sample is predicted for the ith anchor,
1 in the case of positive samples and 0 in the case of negative samples reg Number of anchors, t i To predict the regression parameters corresponding to the ith anchor,
the regression parameter of the GTBox corresponding to the ith anchor is taken as the regression parameter of the GTBox;
the relevant Fast-RCNN loss function is designed as:
L(p，u，t u ，v)＝L cls (p，u)+λ[u≥1]L loc (t u ，v)
wherein, t u V is a bounding box regression parameter corresponding to the real target for predicting the regression parameter corresponding to the category u;
the relevant Mask loss function is designed as:
L(m，n)＝L BCE (m，n)
wherein m is Mask corresponding to the prediction category, and n is GT Mask.
7. The method for detecting and evaluating the defects on the processed surface of the composite material according to any one of claims 1 to 6, wherein in the fourth step, the model test process is preferably designed as follows:
obtaining a plurality of feature maps through a backbone and an FPN, generating corresponding explosals of each feature map through an RPN, and mapping the explosals onto the corresponding feature maps to obtain the explosals feature maps; and then, obtaining a prediction type and a related offset corresponding to the propsall through RoIAlign, two full-connection layers and two parallel full-connection layers, mapping the offset propsals output by the Fast-RCNN network back to a characteristic diagram, changing the size through RoIAlign, inputting the result to a Mask branch, and selecting a Mask corresponding to the target prediction type to map back to the original image.
8. The method for detecting and evaluating the defects on the processed surface of the composite material as claimed in any one of claims 1 to 7, wherein in the fifth step, the defect evaluation score criterion is preferably selected from the group consisting of mechanical properties, physical properties, chemical properties, service life and the like, wherein the mechanical properties further include yield strength, shear modulus, section shrinkage and the like, the physical properties further include electrical resistivity, thermal conductivity, refractive index and the like, the chemical properties further include corrosion resistance, oxidation resistance and the like, and the parameters are proportionally adjusted by weighting coefficients.
9. The method for detecting and evaluating the defects on the processed surface of the composite material according to any one of claims 1 to 8, wherein in the fifth step, the mapping relation between the evaluation index and the evaluation score is established preferably through an artificial neural network, wherein the defect types are represented by weighting corresponding parameters of different types of defects.
10. A composite material processing surface defect detection and evaluation system based on deep learning is characterized by comprising:
an image acquisition and labeling module for capturing images of the composite material machined surface, summarizing the acquired images into an image set and performing defect labeling, thereby forming an image data set;
the image data set dividing and data enhancing module is used for dividing the formed image data set into a training set and a testing set, respectively training a model and the testing model, and simultaneously carrying out image data enhancing processing on the divided training set so as to enlarge the scale of the training set;
the defect detection model training module is used for inputting the obtained training set into the deep learning model to carry out model training;
the defect detection model testing module is used for inputting the obtained testing set into a trained deep learning model for testing to obtain information such as defect types, defect positions, defect depths, defect areas, defect area ratios, length and width of a minimum circumscribed rectangle of a defect area and the like in an image so as to obtain corresponding defect evaluation index values;
and the defect evaluation module is used for forming a mapping relation between the evaluation index and the evaluation score based on the acquired defect evaluation index value and in combination with a preset defect evaluation score criterion, so that the whole detection and evaluation process is completed.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211311389.2A CN115661071A (en) | 2022-10-25 | 2022-10-25 | Composite material processing surface defect detection and evaluation method based on deep learning |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211311389.2A CN115661071A (en) | 2022-10-25 | 2022-10-25 | Composite material processing surface defect detection and evaluation method based on deep learning |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN115661071A true CN115661071A (en) | 2023-01-31 |
Family
ID=84992256
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202211311389.2A Pending CN115661071A (en) | 2022-10-25 | 2022-10-25 | Composite material processing surface defect detection and evaluation method based on deep learning |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115661071A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117173461A (en) * | 2023-08-29 | 2023-12-05 | 湖北盛林生物工程有限公司 | Multi-visual task filling container defect detection method, system and medium |
| CN117274257A (en) * | 2023-11-21 | 2023-12-22 | 山西集智数据服务有限公司 | Automatic classification system for book looks based on machine vision |
-
2022
- 2022-10-25 CN CN202211311389.2A patent/CN115661071A/en active Pending
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117173461A (en) * | 2023-08-29 | 2023-12-05 | 湖北盛林生物工程有限公司 | Multi-visual task filling container defect detection method, system and medium |
| CN117274257A (en) * | 2023-11-21 | 2023-12-22 | 山西集智数据服务有限公司 | Automatic classification system for book looks based on machine vision |
| CN117274257B (en) * | 2023-11-21 | 2024-01-23 | 山西集智数据服务有限公司 | Automatic classification system for book looks based on machine vision |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN115661071A (en) | Composite material processing surface defect detection and evaluation method based on deep learning | |
| CN113610778B (en) | Bridge surface crack detection method and system based on semantic segmentation | |
| CN117291913B (en) | Apparent crack measuring method for hydraulic concrete structure | |
| CN116883393B (en) | Metal surface defect detection method based on anchor frame-free target detection algorithm | |
| Xue et al. | A high efficiency deep learning method for the x-ray image defect detection of casting parts | |
| Zhang et al. | Intelligent defect detection method for additive manufactured lattice structures based on a modified YOLOv3 model | |
| CN115456938A (en) | Metal part crack detection method based on deep learning and ultrasonic infrared image | |
| Chen et al. | An intelligent vision recognition method based on deep learning for pointer meters | |
| Fan et al. | Application of YOLOv5 neural network based on improved attention mechanism in recognition of Thangka image defects | |
| CN116863271A (en) | Lightweight infrared flame detection method based on improved YOLO V5 | |
| Xin et al. | A fine extraction algorithm for image-based surface cracks in underwater dams | |
| Warren et al. | Grain and grain boundary segmentation using machine learning with real and generated datasets | |
| Cheng et al. | Automatic ROI recognition and parameters selection for digital image correlation in measuring structures with complex shapes | |
| CN117635628A (en) | Sea-land segmentation method based on context attention and boundary perception guidance | |
| Ding et al. | A bearing surface defect detection method based on multi-attention mechanism Yolov8 | |
| CN116205918B (en) | Multi-mode fusion semiconductor detection method, device and medium based on graph convolution | |
| Lv et al. | FLCNet: faster and lighter cross-scale feature aggregation network for lead bar surface defect detection | |
| CN115471724A (en) | Fine-grained fish epidemic disease identification fusion algorithm based on self-adaptive normalization | |
| CN114463628A (en) | Deep learning remote sensing image ship target identification method based on threshold value constraint | |
| CN111881932B (en) | FasterRCNN target detection algorithm for military aircraft | |
| Dahai et al. | A lightweight convolutional neural network for recognition and classification for Si3N4 chip substrate surface defects | |
| CN113343932A (en) | River course floater detection algorithm under static camera | |
| Pan et al. | Embedded U-Net: combines multiple feature fusion encode and subpixel reconstruction for microcracks salient object detection | |
| CN113762359B (en) | Deep learning model evaluation system and method for RD time-frequency data | |
| Li | Image recognition method of surface defects of prefabricated concrete members in prefabricated building |
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 |