CN114565576A - DMTL surface defect detection method, device and terminal - Google Patents
DMTL surface defect detection method, device and terminal Download PDFInfo
- Publication number
- CN114565576A CN114565576A CN202210176036.XA CN202210176036A CN114565576A CN 114565576 A CN114565576 A CN 114565576A CN 202210176036 A CN202210176036 A CN 202210176036A CN 114565576 A CN114565576 A CN 114565576A
- Authority
- CN
- China
- Prior art keywords
- defect
- dmtl
- image
- deep learning
- learning target
- 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
- 230000007547 defect Effects 0.000 title claims abstract description 229
- 238000001514 detection method Methods 0.000 title claims abstract description 143
- 238000013135 deep learning Methods 0.000 claims abstract description 50
- 238000000034 method Methods 0.000 claims abstract description 24
- 238000012549 training Methods 0.000 claims description 21
- 238000004519 manufacturing process Methods 0.000 claims description 7
- 238000003860 storage Methods 0.000 claims description 5
- 238000004140 cleaning Methods 0.000 claims description 3
- 230000004807 localization Effects 0.000 claims description 3
- 230000006870 function Effects 0.000 abstract description 8
- 239000000446 fuel Substances 0.000 abstract description 7
- 238000002372 labelling Methods 0.000 abstract description 7
- 238000012545 processing Methods 0.000 abstract description 7
- 238000011084 recovery Methods 0.000 abstract description 6
- 238000013461 design Methods 0.000 abstract description 2
- 239000010410 layer Substances 0.000 description 28
- 230000008569 process Effects 0.000 description 11
- 238000010586 diagram Methods 0.000 description 7
- 238000005457 optimization Methods 0.000 description 4
- 238000004891 communication Methods 0.000 description 3
- 238000004590 computer program Methods 0.000 description 2
- 238000007789 sealing Methods 0.000 description 2
- 238000013500 data storage Methods 0.000 description 1
- 238000003745 diagnosis Methods 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 238000012854 evaluation process Methods 0.000 description 1
- 239000002737 fuel gas Substances 0.000 description 1
- 230000002068 genetic effect Effects 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 230000001788 irregular Effects 0.000 description 1
- 239000002346 layers by function Substances 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 239000002245 particle Substances 0.000 description 1
- 230000002787 reinforcement Effects 0.000 description 1
- 238000009877 rendering Methods 0.000 description 1
- 238000012546 transfer Methods 0.000 description 1
- 230000000007 visual effect Effects 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/0002—Inspection of images, e.g. flaw detection
- G06T7/0004—Industrial image inspection
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02P—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN THE PRODUCTION OR PROCESSING OF GOODS
- Y02P90/00—Enabling technologies with a potential contribution to greenhouse gas [GHG] emissions mitigation
- Y02P90/30—Computing systems specially adapted for manufacturing
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Computation (AREA)
- General Engineering & Computer Science (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- Bioinformatics & Computational Biology (AREA)
- Evolutionary Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Biomedical Technology (AREA)
- Health & Medical Sciences (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Quality & Reliability (AREA)
- Investigating Materials By The Use Of Optical Means Adapted For Particular Applications (AREA)
- Image Analysis (AREA)
Abstract
The application discloses a DMTL surface defect detection method, a DMTL surface defect detection device and a DMTL surface defect detection terminal, and belongs to the field of design of fuel vapor recovery systems. In the method, after a terminal acquires a DMTL surface image and inputs the image into a deep learning target detection model, a surface defect detection result output by the deep learning target detection model is obtained, and then defect labeling is carried out on the DMTL surface according to the surface defect detection result. The defect size, position and type of the DMTL are detected through the defect detection function of the deep learning target detection model, the task which is difficult to complete by the traditional image processing algorithm is completed, and manual work is replaced; different from the mode of manually marking the pictures, the defect mark is generated by detecting the unmarked pictures through the trained model, so that a large amount of manual labor is saved.
Description
Technical Field
The invention relates to the field of design of fuel vapor recovery systems, in particular to a DMTL surface defect detection method, a device and a terminal.
Background
The fuel emission leakage Diagnosis Module (DMTL) is added on the basis of a fuel vapor recovery system in order to meet the requirements of the national six-emission regulations, and can diagnose whether fuel gas leaks. The DMTL comprises a plurality of important functional hardware interfaces, such as an air outlet end face, an air inlet, a screw port, a wiring harness plug face and the like, wherein the functional faces have extremely strict detection requirements in the production process, and the defects of flash, deformation, defect and the like of the functional faces can cause the problems of non-sealing of a sealing port and poor contact of a plug; meanwhile, the problems of scratches, pits, deformation and the like of the non-functional surface can cause customer complaints and the like to influence the quality image of the company product.
The DMTL surface is made of plastic and comprises various complex parts with three-dimensional geometric shapes, the brightness of the plastic surface is completely different at different angles under the light of a light source, therefore, the detection of the defects of any position and irregularity of the surface can hardly be finished under the traditional visual logic, the plastic surface has very uneven brightness under the light, and the defects with complex shapes are in different states at different positions.
Disclosure of Invention
The invention provides a DMTL surface defect detection method, a device and a terminal, which can solve the problem that good defect detection cannot be carried out on an irregular DMTL surface in the related technology. The technical scheme is as follows:
in one aspect, the present invention provides a DMTL surface defect detection method, for a terminal, the method comprising:
acquiring a DMTL surface image;
inputting the DMTL surface image into a deep learning target detection model to obtain a surface defect detection result output by the deep learning target detection model;
and marking the defects on the DMTL surface according to the surface defect detection result.
In another aspect, the present invention provides a DMTL surface defect detection apparatus for a terminal, the apparatus comprising:
the image acquisition module is used for acquiring a DMTL surface image;
the defect detection module is used for inputting the DMTL surface image into a deep learning target detection model to obtain a surface defect detection result output by the deep learning target detection model;
and the defect marking module is used for marking the defects on the DMTL surface according to the surface defect detection result.
In another aspect, the present invention provides a terminal comprising a processor and a memory; the memory stores at least one instruction for execution by the processor to implement the DMTL surface defect detection method of the above aspect.
In another aspect, a computer-readable storage medium is provided that stores at least one instruction for execution by a processor to implement the DMTL surface defect detection method of the above aspect.
By adopting the DMTL surface defect detection method provided by the invention, after the terminal acquires the DMTL surface image and inputs the image into the deep learning target detection model, the surface defect detection result output by the deep learning target detection model is obtained, and then the defect marking is carried out on the DMTL surface according to the surface defect detection result. The defect size, position and type of the DMTL are detected through the defect detection function of the deep learning target detection model, the task which is difficult to complete by the traditional image processing algorithm is completed, and manual work is replaced; different from the mode of manually marking the pictures, the defect mark is generated by detecting the unmarked pictures through the trained model, so that a large amount of manual labor is saved.
Drawings
FIG. 1 illustrates a flow chart of a DMTL surface defect detection method shown in an exemplary embodiment of the present application;
FIG. 2 illustrates a flow chart of a DMTL surface defect detection method shown in another exemplary embodiment of the present application;
FIG. 3 illustrates a schematic structural diagram of a deep learning object detection model according to an exemplary embodiment of the present application;
FIG. 4 shows a schematic diagram of a training process for a deep learning object detection model;
FIG. 5 shows a schematic diagram of a hyperparametric search flow of a Bayesian optimization algorithm;
FIG. 6 shows a block diagram of a DMTL surface defect detection apparatus provided in one embodiment of the present application;
fig. 7 is a block diagram illustrating a structure of a terminal according to an exemplary embodiment of the present application.
Detailed Description
To make the objects, technical solutions and advantages of the present invention more apparent, embodiments of the present invention will be described in detail with reference to the accompanying drawings.
Reference herein to "a plurality" means two or more. "and/or" describes the association relationship of the associated objects, meaning that there may be three relationships, e.g., a and/or B, which may mean: a exists alone, A and B exist simultaneously, and B exists alone. The character "/" generally indicates that the former and latter associated objects are in an "or" relationship.
It should be noted that the DMTL surface defect detection method of the present invention is applicable to a fuel vapor recovery system, and the DMTL surface defect detection method may be integrated into a terminal detection device to be associated with the fuel vapor recovery system, and in the operation process of the fuel vapor recovery system, the DMTL surface defect detection method may be an application program independent of a terminal.
Example 1
Referring to fig. 1, a flow chart of a DMTL surface defect detection method according to an exemplary embodiment of the present application is shown, the method including:
In the invention, the DMTL surface defect detection method program executed on the terminal can realize the real-time provision of DMTL surface defect detection service for DMTL to be detected.
In a possible embodiment, the terminal is provided with at least one image acquisition device, which may be mobile or stationary. The image acquisition device is used for acquiring a DMTL surface image.
In one example, the detection items on the DMTL surface may be a screw port, an air outlet, an air inlet, a wiring harness plug, and the like, which is not limited in this embodiment. A separate image capturing device may be disposed at a corresponding item position, or sequential capturing may be performed by a mobile image capturing device, which is not limited in this application.
And 102, inputting the DMTL surface image into a deep learning target detection model to obtain a surface defect detection result output by the deep learning target detection model.
The method is used for realizing the DMTL surface defect detection function based on a deep learning target detection model.
In a possible implementation manner, the terminal inputs the acquired DMTL surface image into the deep learning target detection model, and the deep learning target detection model performs defect detection on the DMTL surface image to obtain a surface defect detection result output by the deep learning target detection model.
Optionally, the surface defect detection result may characterize the defect region location, size, and defect type of the DMTL surface image, wherein the defect type is, for example, scratch, flash, dent, deformation, and the like.
In an illustrative example, a screw opening image of the surface of the DMTL is collected, the screw opening image is input into a deep learning target detection model, and a screw opening image defect detection result output by the deep learning target detection model is obtained, wherein the screw opening image defect detection result represents the position, the size and the defect type of a defect area with the screw opening image.
And 103, marking the defects on the DMTL surface according to the surface defect detection result.
Furthermore, according to the obtained surface defect detection result, the position of a defect area is determined on the surface of the DMTL, then defect marking is carried out according to the size (for example, the defect area is marked through a rectangular frame), in addition, the defect category is marked at the marking position, and therefore the complete process of defect marking is completed.
In summary, with the DMTL surface defect detection method provided by the present invention, after the terminal obtains the DMTL surface image and inputs the image to the deep learning target detection model, the surface defect detection result output by the deep learning target detection model is obtained, and then the defect labeling is performed on the DMTL surface according to the surface defect detection result. The defect size, position and type of the DMTL are detected through the defect detection function of the deep learning target detection model, the task which is difficult to complete by the traditional image processing algorithm is completed, and manual work is replaced; different from the mode of manually marking the pictures, the defect mark is generated by detecting the unmarked pictures through the trained model, so that a large amount of manual labor is saved.
Example 2
Referring to fig. 2, a flow chart of a DMTL surface defect detection method according to another exemplary embodiment of the present application is shown, the method comprising:
The step 101 may be referred to in the implementation manner of this step, and this embodiment is not described herein again.
In one possible implementation, as shown in fig. 3, the deep learning object detection model in the above embodiment includes a defect localization layer and a defect classification layer.
Taking the detection items of the DMTL surface as a screw hole, an air outlet, an air inlet and a wiring harness plug as an example, the step 202 explains that the screw hole image is input into a defect positioning layer to obtain screw hole defect positioning information, and the screw hole defect positioning information comprises the position and the size of each defect area in the screw hole image; inputting the air outlet image into a defect positioning layer to obtain air outlet defect positioning information, wherein the air outlet defect positioning information comprises the position and the size of each defect area in the air outlet image; inputting the air inlet image into a defect positioning layer to obtain air inlet defect positioning information, wherein the air inlet defect positioning information comprises the position and the size of each defect area in the air inlet image; and inputting the wiring harness plug image into the defect positioning layer to obtain wiring harness plug defect positioning information, wherein the wiring harness plug defect positioning information comprises the position and the size of each defect area in the wiring harness plug image.
Taking the detection items of the DMTL surface as a screw hole, an air outlet, an air inlet and a wiring harness plug as an example, the step 203 explains that the screw hole image is input into a defect classification layer to obtain screw hole defect classification information, and the screw hole defect classification information comprises defect categories of defect areas in the screw hole image; inputting the air outlet image into a defect classification layer to obtain air outlet defect classification information, wherein the air outlet defect classification information comprises defect types of all defect areas in the air outlet image; inputting the air inlet image into a defect classification layer to obtain air inlet defect classification information, wherein the air inlet defect classification information comprises defect types of all defect areas in the air inlet image; and inputting the wiring harness plug image into a defect classification layer to obtain wiring harness plug defect classification information, wherein the wiring harness plug defect classification information comprises defect types of all defect areas in the wiring harness plug image.
And 204, integrating the defect positioning information and the defect classification information to obtain a surface defect detection result corresponding to the DMTL surface image.
Taking the detection items of the DMTL surface as a screw port, an air outlet, an air inlet and a wiring harness plug as an example, the step 204 explains that the screw port defect positioning information and the screw port defect classification information are integrated to obtain a screw port surface defect detection result; integrating the air outlet defect positioning information and the air outlet defect classification information to obtain an air outlet surface defect detection result; integrating air inlet defect positioning information and air inlet defect classification information to obtain an air inlet surface defect detection result; and integrating the wiring harness plug defect positioning information and the wiring harness plug defect classification information to obtain a wiring harness plug surface defect detection result.
And step 205, marking the defects on the DMTL surface according to the surface defect detection result.
Taking the detection items of the DMTL surface as a screw port, an air outlet, an air inlet and a wiring harness plug as examples, the step 205 is explained as marking the defects on the surface of the screw port according to the detection result of the defects on the surface of the screw port; marking the defects on the surface of the air outlet according to the detection result of the defects on the surface of the air outlet; marking the surface of the air inlet with defects according to the detection result of the surface defects of the air inlet; and marking the defects on the surface of the wire harness plug according to the detection result of the defects on the surface of the wire harness plug.
In the embodiment of the application, the functional layer structure of the deep learning target detection model is further disclosed, the defect detection of the model is further explained, the defect position and the size of the graph can be obtained through the defect positioning layer, the defect category of the graph is obtained through the defect classification layer, and therefore the surface defect detection result is output.
Example 3
As shown in fig. 4, the embodiment of the present application describes a training process of a deep learning target detection model. I.e. before step 101 or step 201, the method further comprises:
In the model training process, the result comparison is carried out through the labeled sample image, and the labeling mode can be used for indicating the defect position through the rectangular frame.
The image labeling may include problems such as misspelling, misclassification, and unsatisfactory labeling position and size, and the initial defect labeling needs to be analyzed and cleaned to obtain an initial data set, so as to meet the requirements.
And 403, performing data enhancement on the initial data set in an off-line and/or on-line mode to generate a target data set corresponding to the sample image.
In one possible implementation mode, in order to improve the performance of the deep learning target detection model, offline, online or offline and online mixed data enhancement can be performed, and the offline data enhancement means that a target data set corresponding to a sample image is generated after data is enhanced in a fixed manner; the online data enhancement means that the data is enhanced according to a certain rule in the training process.
And step 404, acquiring network parameters and image adjustment proportion parameters of the defect positioning layer and the defect classification layer based on a hyper-parameter searching mode.
In the embodiment of the application, the hyper-parameter search mode is schematically described by using a bayesian optimization algorithm, including but not limited to a bayesian optimization algorithm, a genetic algorithm, a differential evolution algorithm, a particle swarm algorithm and a reinforcement learning algorithm. As shown in FIG. 5, a hyperparametric search process using a Bayesian optimization algorithm is schematically illustrated.
Wherein, the acquired new hyper-parameters are network parameters and image adjusting proportion parameters of the defect positioning layer and the defect classification layer,
The deep learning target detection model comprises a large number of variable learning parameters and network parameters, and scaling parameters are required to be adjusted when an input image is too large.
Furthermore, the hyperparametric search process may add the detection time as a penalty term or a target term in case there is a requirement for the detection time.
And 406, performing iterative training on the defect positioning layer and the defect classification layer according to the sample image.
And 407, responding to the iteration times meeting the preset number, and outputting a sample image model training result.
And step 408, comparing the sample image model training result with the target data set, and outputting the optimal deep learning target detection model.
Further, by steps 406 to 408, the best model is selected and searched within a preset number of iterations (i.e., the model evaluation process), and the optimal deep learning target detection model is finally output.
And step 409, adding artificial rules to the deep learning target detection model based on the model output condition.
The manual rule adding means that false defects such as the defect size is too small and the defect confidence coefficient is too low which are not in a specified area can be detected in the actual model deployment process, and the false alarm rate of the defect can be greatly reduced by manually setting the rule.
And adding artificial rules to the deep learning target detection model based on the output condition of the model, wherein if the rules are that the threshold value of the output class probability is selected, the logic relation of the defects is judged, and the like.
And step 410, carrying out algorithm field deployment on the deep learning target detection model according to the production line detection requirement.
The deep learning target detection model comprises a plurality of artificial rule algorithms which are deployed on a production line, and can use modes such as local industrial personal computer deployment, network server deployment and the like to ensure that the real-time production line detection requirements of the production line are met.
Referring to fig. 6, a block diagram of a DMTL surface defect detection apparatus according to an embodiment of the present application is shown. The apparatus may be implemented as all or a portion of the terminal in software, hardware, or a combination of both. The device includes:
the image acquisition module 601 is used for acquiring a DMTL surface image;
a defect detection module 602, configured to input the DMTL surface image into a deep learning target detection model to obtain a surface defect detection result output by the deep learning target detection model;
and a defect labeling module 603, configured to label defects on the DMTL surface according to the surface defect detection result.
Optionally, the deep learning target detection model includes a defect localization layer and a defect classification layer.
Optionally, the defect detecting module 602 includes:
the first detection submodule is used for inputting the DMTL surface image into the defect positioning layer to obtain defect positioning information, and the defect positioning information comprises the position and the size of each defect area in the DMTL surface image;
the second detection submodule is used for inputting the DMTL surface image into the defect classification layer to obtain defect classification information, and the defect classification information comprises defect categories of all defect areas in the DMTL surface image;
and the third detection submodule is used for integrating the defect positioning information and the defect classification information to obtain a surface defect detection result corresponding to the DMTL surface image.
Optionally, the apparatus further comprises:
the first training module is used for acquiring initial defect labels of the sample images;
the second training module is used for analyzing and cleaning the initial defect labels to obtain an initial data set;
the third training module is used for performing data enhancement on the initial data set in an off-line and/or on-line mode to generate a target data set corresponding to the sample image;
the fourth training module is used for acquiring network parameters and image adjustment proportion parameters of the defect positioning layer and the defect classification layer based on a hyper-parameter searching mode;
the fifth training module is used for searching and configuring the learning rate and the learning interval of the network parameters and the image scaling of the image adjustment scale parameters;
the sixth training module is used for performing iterative training on the defect positioning layer and the defect classification layer according to the sample image;
the seventh training module is used for responding to the fact that the iteration times meet the preset number and outputting a sample image model training result;
and the eighth training module is used for comparing the sample image model training result with the target data set and outputting an optimal deep learning target detection model.
Optionally, the apparatus further comprises:
and the rule adding module is used for adding artificial rules to the deep learning target detection model according to the false alarm information.
Optionally, the apparatus further comprises:
and the field deployment module is used for carrying out algorithm field deployment on the deep learning target detection model according to the production line detection requirement.
Referring to fig. 7, a block diagram of a terminal 700 according to an exemplary embodiment of the present application is shown. The terminal 700 may be an electronic device installed and running an application, such as a smart phone, a tablet computer, an electronic book, a portable personal computer, or the like. The terminal 700 in the present application may include one or more of the following components: a processor 710, a memory 720, and a screen 730.
The Memory 720 may include a Random Access Memory (RAM) or a Read-Only Memory (Read-Only Memory). Optionally, the memory 720 includes a non-transitory computer-readable medium. The memory 720 may be used to store instructions, programs, code sets, or instruction sets. The memory 720 may include a program storage area and a data storage area, wherein the program storage area may store instructions for implementing an operating system, instructions for implementing at least one function, instructions for implementing the various method embodiments described above, and the like, and the type of operating system is not limited. The storage data area may also store data created by terminal 700 during use, and the like.
The screen 730 may be a touch screen or a keyboard and mouse operated screen, and is generally disposed on a front panel of the terminal 700.
The invention also provides a computer readable medium storing at least one instruction, the at least one instruction being loaded and executed by the processor to implement the DMTL surface defect detection method according to the above embodiments.
The present invention also provides a computer program product storing at least one instruction, the at least one instruction being loaded and executed by the processor to implement the DMTL surface defect detection method according to the various embodiments described above.
Those skilled in the art will recognize that, in one or more of the examples described above, the functions described in this invention may be implemented in hardware, software, firmware, or any combination thereof. When implemented in software, the functions may be stored on or transmitted over as one or more instructions or code on a computer-readable medium. Computer-readable media includes both computer storage media and communication media including any medium that facilitates transfer of a computer program from one place to another. A storage media may be any available media that can be accessed by a general purpose or special purpose computer.
The above description is only exemplary of the present application and should not be taken as limiting, as any modification, equivalent replacement, or improvement made within the spirit and principle of the present application should be included in the protection scope of the present application.
Claims (9)
1. A DMTL surface defect detection method for a terminal, the method comprising:
acquiring a DMTL surface image;
inputting the DMTL surface image into a deep learning target detection model to obtain a surface defect detection result output by the deep learning target detection model;
and marking the defects on the DMTL surface according to the surface defect detection result.
2. The method of claim 1, wherein the deep-learning object detection model comprises a defect localization layer and a defect classification layer.
3. The method of claim 2, wherein the inputting the DMTL surface image into a deep learning target detection model to obtain a surface defect detection result output by the deep learning target detection model comprises:
inputting the DMTL surface image into the defect positioning layer to obtain defect positioning information, wherein the defect positioning information comprises the position and the size of each defect area in the DMTL surface image;
inputting the DMTL surface image into the defect classification layer to obtain defect classification information, wherein the defect classification information comprises defect categories of all defect areas in the DMTL surface image;
and integrating the defect positioning information and the defect classification information to obtain a surface defect detection result corresponding to the DMTL surface image.
4. The method of claim 2, further comprising:
acquiring initial defect labels of sample images;
analyzing and cleaning the initial defect labels to obtain an initial data set;
performing data enhancement on the initial data set in an off-line and/or on-line mode to generate a target data set corresponding to the sample image;
acquiring network parameters and image adjustment proportion parameters of the defect positioning layer and the defect classification layer based on a super-parameter searching mode;
searching and configuring the learning rate and the learning interval of the network parameters and the image scaling of the image adjustment scale parameter;
performing iterative training on the defect positioning layer and the defect classification layer according to the sample image;
responding to the iteration times meeting the preset number, and outputting a training result of the sample image model;
and comparing the sample image model training result with the target data set, and outputting an optimal deep learning target detection model.
5. The method of claim 4, further comprising:
and adding artificial rules to the deep learning target detection model based on the model output condition.
6. The method of claim 4, further comprising:
and carrying out algorithm field deployment on the deep learning target detection model according to the detection requirement of a production line.
7. A DMTL surface defect detection apparatus, wherein the apparatus is adapted for termination, the apparatus comprising:
the image acquisition module is used for acquiring a DMTL surface image;
the defect detection module is used for inputting the DMTL surface image into a deep learning target detection model to obtain a surface defect detection result output by the deep learning target detection model;
and the defect marking module is used for marking the defects on the DMTL surface according to the surface defect detection result.
8. A terminal, characterized in that the terminal comprises a processor and a memory; the memory stores at least one instruction for execution by the processor to implement the DMTL surface defect detection method of any of claims 1 to 6.
9. A computer-readable storage medium having stored thereon at least one instruction for execution by a processor to perform the DMTL surface defect detection method of any of claims 1 to 6.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210176036.XA CN114565576A (en) | 2022-02-25 | 2022-02-25 | DMTL surface defect detection method, device and terminal |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210176036.XA CN114565576A (en) | 2022-02-25 | 2022-02-25 | DMTL surface defect detection method, device and terminal |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN114565576A true CN114565576A (en) | 2022-05-31 |
Family
ID=81715555
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210176036.XA Pending CN114565576A (en) | 2022-02-25 | 2022-02-25 | DMTL surface defect detection method, device and terminal |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114565576A (en) |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108257114A (en) * | 2017-12-29 | 2018-07-06 | 天津市万贸科技有限公司 | A kind of transmission facility defect inspection method based on deep learning |
| CN111862067A (en) * | 2020-07-28 | 2020-10-30 | 中山佳维电子有限公司 | Welding defect detection method and device, electronic equipment and storage medium |
| WO2021022933A1 (en) * | 2019-08-06 | 2021-02-11 | 平安科技（深圳）有限公司 | Method and device for multitask prediction, electronic device, and storage medium |
| CN113780342A (en) * | 2021-08-04 | 2021-12-10 | 杭州国辰机器人科技有限公司 | Intelligent detection method and device based on self-supervision pre-training and robot |
-
2022
- 2022-02-25 CN CN202210176036.XA patent/CN114565576A/en active Pending
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108257114A (en) * | 2017-12-29 | 2018-07-06 | 天津市万贸科技有限公司 | A kind of transmission facility defect inspection method based on deep learning |
| WO2021022933A1 (en) * | 2019-08-06 | 2021-02-11 | 平安科技（深圳）有限公司 | Method and device for multitask prediction, electronic device, and storage medium |
| CN111862067A (en) * | 2020-07-28 | 2020-10-30 | 中山佳维电子有限公司 | Welding defect detection method and device, electronic equipment and storage medium |
| CN113780342A (en) * | 2021-08-04 | 2021-12-10 | 杭州国辰机器人科技有限公司 | Intelligent detection method and device based on self-supervision pre-training and robot |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110148130B (en) | Method and device for detecting part defects | |
| US9349076B1 (en) | Template-based target object detection in an image | |
| CN111310826B (en) | Method and device for detecting labeling abnormality of sample set and electronic equipment | |
| CN110264445A (en) | The screen printing of battery quality determining method of piecemeal template matching combining form processing | |
| CN111666940B (en) | Chat screenshot content processing method and device, electronic equipment and readable storage medium | |
| CN114549993B (en) | Method, system and device for grading line segment image in experiment and readable storage medium | |
| CN116485735A (en) | Defect detection method, defect detection device, electronic equipment and computer readable storage medium | |
| CN117392042A (en) | Defect detection method, defect detection apparatus, and storage medium | |
| CN114255223A (en) | Deep learning-based method and equipment for detecting surface defects of two-stage bathroom ceramics | |
| CN113838015B (en) | Electrical product appearance defect detection method based on network cooperation | |
| CN115984662A (en) | Multi-mode data pre-training and recognition method, device, equipment and medium | |
| CN113821296B (en) | Visual interface generation method, electronic equipment and storage medium | |
| CN110910414A (en) | Image contour generation method, image labeling method, electronic device and storage medium | |
| CN115546824B (en) | Taboo picture identification method, apparatus and storage medium | |
| CN114565576A (en) | DMTL surface defect detection method, device and terminal | |
| CN111062388A (en) | Advertisement character recognition method, system, medium and device based on deep learning | |
| JP2010140444A (en) | Inspection system and method of controlling same, and computer program | |
| CN116992519A (en) | Fire-fighting construction drawing marking method and device, electronic equipment and storage medium | |
| CN115829925A (en) | Appearance defect detection method and device, computer equipment and storage medium | |
| CN115731208A (en) | High-precision defect detection method, device and equipment for metal complex surface and storage medium | |
| CN116977249A (en) | Defect detection method, model training method and device | |
| CN113744392A (en) | Three-dimensional model library construction method, device, equipment and medium | |
| CN111727108B (en) | Method, device and system for controlling robot and storage medium | |
| CN112131418A (en) | Target labeling method, target labeling device and computer-readable storage medium | |
| Chen | Development of image recognition system for steel defects detection |
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 |