CN113674277A - Unsupervised domain adaptive surface defect region segmentation method and device and electronic equipment - Google Patents
Unsupervised domain adaptive surface defect region segmentation method and device and electronic equipment Download PDFInfo
- Publication number
- CN113674277A CN113674277A CN202111230003.0A CN202111230003A CN113674277A CN 113674277 A CN113674277 A CN 113674277A CN 202111230003 A CN202111230003 A CN 202111230003A CN 113674277 A CN113674277 A CN 113674277A
- Authority
- CN
- China
- Prior art keywords
- domain image
- image
- defect region
- target
- region segmentation
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
- 230000007547 defect Effects 0.000 title claims abstract description 69
- 230000011218 segmentation Effects 0.000 title claims abstract description 60
- 238000000034 method Methods 0.000 title claims abstract description 45
- 230000003044 adaptive effect Effects 0.000 title abstract description 8
- 238000009826 distribution Methods 0.000 claims abstract description 40
- 238000001228 spectrum Methods 0.000 claims description 77
- 238000001514 detection method Methods 0.000 abstract description 7
- 230000009286 beneficial effect Effects 0.000 abstract description 4
- 238000010586 diagram Methods 0.000 description 5
- 229910000976 Electrical steel Inorganic materials 0.000 description 4
- 229910000831 Steel Inorganic materials 0.000 description 4
- 238000004519 manufacturing process Methods 0.000 description 4
- 239000010959 steel Substances 0.000 description 4
- 230000009466 transformation Effects 0.000 description 4
- 238000005286 illumination Methods 0.000 description 3
- 238000007689 inspection Methods 0.000 description 2
- 230000000007 visual effect Effects 0.000 description 2
- 230000006978 adaptation Effects 0.000 description 1
- 230000007797 corrosion Effects 0.000 description 1
- 238000005260 corrosion Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 238000009776 industrial production Methods 0.000 description 1
- 238000010801 machine learning Methods 0.000 description 1
- 239000002994 raw material Substances 0.000 description 1
- 238000005096 rolling process Methods 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/0002—Inspection of images, e.g. flaw detection
- G06T7/0004—Industrial image inspection
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N20/00—Machine learning
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T5/00—Image enhancement or restoration
- G06T5/10—Image enhancement or restoration using non-spatial domain filtering
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/11—Region-based segmentation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20081—Training; Learning
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30108—Industrial image inspection
- G06T2207/30136—Metal
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Software Systems (AREA)
- Artificial Intelligence (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Medical Informatics (AREA)
- Quality & Reliability (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- Mathematical Physics (AREA)
- Image Analysis (AREA)
Abstract
The application relates to a unsupervised domain adaptive surface defect region segmentation method, a device and electronic equipment, belonging to the technical field of industrial detection; utilizing the trained defect region segmentation model to carry out defect region segmentation processing on the target surface image; the training process of the defect region segmentation model comprises the following steps: acquiring a target domain image belonging to the same application scene as the target surface image; processing the original source domain image with the label according to the target domain image to obtain a source domain image with the data distribution characteristic of the target domain image; and carrying out supervised training on the model by using the source domain image with the target domain image data distribution characteristics to obtain the trained defect region segmentation model. The method and the device for segmenting the image region defects are beneficial to more efficiently improving the performance of the segmentation model under various application scenes, and therefore the accuracy of segmentation of the region defects of the image under specific application scenes is improved.
Description
Technical Field
The application belongs to the technical field of industrial detection, and particularly relates to a unsupervised domain adaptive surface defect region segmentation method and device and electronic equipment.
Background
In the industrial production process, the surface defect detection of related products is required. For example, in the production and processing process of steel, the defects such as holes, scratches, inclusions, scratches, roll marks and the like are easily generated due to the influence of a plurality of factors such as raw materials, rolling equipment, operating techniques of workers and the like, and the appearance of the steel is affected and the properties such as corrosion resistance, wear resistance, fatigue strength and the like are also affected due to the existence of the defects, so that the quality of the steel is seriously reduced.
The surface defect detection described above involves dividing a surface defect region. In the prior art, a coder-decoder network structure similar to U-Net is adopted, a model is trained on a given training data set with a large number of artificially labeled defect regions to obtain a defect region segmentation model, and the trained model is used for processing an acquired surface image to segment and extract the defect regions.
However, the existing related method only considers the condition that the image data distribution of the training data set (source domain) learned by the defect region segmentation model is consistent with the image data distribution of the test scene (target domain) of model inference. In an actual application scene, due to the fact that changes of visual angles, shelters and illumination conditions exist in different scenes, the scene coverage range of a defect region segmentation data set constructed by people is limited, and the source domain data distribution used for model training and the target domain data distribution used for actual application often have differences, so that the performance of a model is seriously reduced. And the target domain data based on practical application is manually labeled, model training (supervised learning) is carried out again based on the labeled data, and a large amount of labeled data is needed, so that the overall efficiency is not high.
The above is only for the purpose of assisting understanding of the technical aspects of the present invention, and does not represent an admission that the above is prior art.
Disclosure of Invention
In order to overcome the problems in the related art at least to a certain extent, the application provides a unsupervised domain adaptive surface defect region segmentation method, a unsupervised domain adaptive surface defect region segmentation device and electronic equipment, and a specific model training mode is adopted, so that the performance of segmenting a model in various application scenes is improved more efficiently.
In order to achieve the purpose, the following technical scheme is adopted in the application:
in a first aspect,
the application provides a surface defect region segmentation method adaptive to an unsupervised domain, which comprises the following steps:
acquiring a target surface image;
utilizing the trained defect region segmentation model to carry out defect region segmentation processing on the target surface image;
the training process of the defect region segmentation model comprises the following steps:
acquiring a target domain image belonging to the same application scene as the target surface image;
processing the original source domain image with the label according to the target domain image to obtain a source domain image with the data distribution characteristic of the target domain image;
and carrying out supervised training on the model by using the source domain image with the target domain image data distribution characteristics to obtain the trained defect region segmentation model.
Optionally, the processing the original source domain image with the label according to the target domain image to obtain the source domain image with the target domain image data distribution characteristic includes:
respectively carrying out Fourier transform on the target domain image and the original source domain image to obtain a first magnitude spectrum and a first phase spectrum corresponding to the original source domain image and obtain a second magnitude spectrum and a second phase spectrum corresponding to the target domain image;
replacing low-frequency part data in the first amplitude spectrum with low-frequency part data in the second amplitude spectrum to obtain a replaced third amplitude spectrum;
performing inverse Fourier transform according to the third magnitude spectrum and the first phase spectrum, and taking an image obtained after inverse transform as the source domain image with the target domain image data distribution characteristic;
the low-frequency part data is spectrogram data with the frequency less than or equal to a preset threshold frequency in the amplitude spectrum.
Optionally, the method further comprises:
determining an application scene;
and determining the value of the preset threshold frequency based on the determined application scene.
Optionally, a value of the preset threshold frequency is one third of an upper limit value of a frequency range of the amplitude spectrum.
Optionally, the replacing the low-frequency part data in the first amplitude spectrum with the low-frequency part data in the second amplitude spectrum to obtain a replaced third amplitude spectrum, specifically:
determining a circular region to be scratched on a second amplitude spectrum by taking the middle point of the amplitude spectrum as the circle center and a preset length as the radius, wherein the preset length is determined based on the preset threshold frequency;
and extracting a spectrogram corresponding to the circular area on the second magnitude spectrum, and correspondingly replacing the spectrogram into the first magnitude spectrum to obtain the third magnitude spectrum.
Optionally, the defect region segmentation model is a PSPnet model or a U-Net model.
In a second aspect of the present invention,
the present application provides an unsupervised domain adapted surface defect region segmentation apparatus, the apparatus comprising:
the acquisition module is used for acquiring a target surface image;
the segmentation processing module is used for carrying out defect region segmentation processing on the target surface image by utilizing the trained defect region segmentation model;
a training module, configured to train the defect region segmentation model, and specifically configured to:
acquiring a target domain image belonging to the same application scene as the target surface image;
processing the original source domain image with the label according to the target domain image to obtain a source domain image with the data distribution characteristic of the target domain image;
and carrying out supervised training on the model by utilizing the source domain image with the target domain image data distribution characteristic.
In a third aspect,
the application provides an electronic device, including:
a memory having an executable program stored thereon;
a processor for executing the executable program in the memory to implement the steps of the method described above.
This application adopts above technical scheme, possesses following beneficial effect at least:
according to the technical scheme, the original source domain image with the label is processed through the target domain image in the model training process to obtain the source domain image with the target domain image data distribution characteristic, the source domain image with the target domain image data distribution characteristic is used for carrying out supervised training on the model, and the specific model training mode is adopted, so that the performance of segmenting the model in various application scenes is improved more efficiently, and the accuracy of segmenting the defect region of the image in the specific application scene is improved.
Additional advantages, objects, and features of the invention will be set forth in part in the description which follows and in part will become apparent to those having ordinary skill in the art upon examination of the following or may be learned from practice of the invention. The objectives and other advantages of the invention will be realized and attained by the structure particularly pointed out in the written description and claims hereof as well as the appended drawings.
Drawings
The accompanying drawings are included to provide a further understanding of the technology or prior art of the present application and are incorporated in and constitute a part of this specification. The drawings expressing the embodiments of the present application are used for explaining the technical solutions of the present application, and should not be construed as limiting the technical solutions of the present application.
Fig. 1 is a schematic flowchart of a method for unsupervised domain adaptive surface defect region segmentation according to an embodiment of the present application;
FIG. 2 is a schematic diagram illustrating a model training process according to an embodiment of the present application;
FIG. 3 is a schematic structural diagram of an unsupervised domain adapted surface defect region segmentation apparatus according to an embodiment of the present application;
fig. 4 is a schematic structural diagram of an electronic device according to an embodiment of the present application.
Detailed Description
In order to make the objects, technical solutions and advantages of the present application more apparent, the technical solutions of the present application will be described in detail below. It is to be understood that the embodiments described are only a few embodiments of the present application and not all embodiments. All other embodiments, which can be derived by a person skilled in the art from the examples given herein without making any creative effort, shall fall within the protection scope of the present application.
As described in the background, in industrial inspection applications, surface defect inspection of related products is required. The existing related method only considers the condition that the image data distribution of a training data set (source domain) learned by a defect region segmentation model is consistent with the image data distribution of a test scene (target domain) of model inference.
In an actual application scene, due to the fact that changes of visual angles, shelters and illumination conditions exist in different scenes, the scene coverage range of a defect region segmentation data set constructed by people is limited, and the source domain data distribution used for model training and the target domain data distribution used for actual application often have differences, so that the performance of a model is seriously reduced. And the target domain data based on practical application is manually labeled, model training (supervised learning) is carried out again based on the labeled data, and a large amount of labeled data is needed, so that the overall efficiency is not high.
In view of the above, the present application provides a surface defect region segmentation method without supervision domain adaptation. As shown in fig. 1, in one embodiment, the method includes:
step S210, acquiring a target surface image;
for example, the application scenario of the embodiment is a silicon steel strip production scenario, in which surface defect detection needs to be performed on the silicon steel strip, a camera device is specifically configured in a production site, and a surface image (i.e., a target surface image) of the silicon steel strip to be input into the detection processing system is obtained by shooting through the camera device.
Then, as shown in fig. 1, step S220 is performed to perform a defect region segmentation process on the target surface image by using the trained defect region segmentation model, so as to segment the surface defect region.
For example, the defect region segmentation model may be a PSPnet model or a U-Net model (or a model using a coder-decoder network structure similar to U-Net), and the specific structures of these models may be found in the prior art, which is not described herein again.
Unlike the prior art, in this embodiment, as shown in fig. 1, the training process of the defect region segmentation model includes:
step S110, acquiring a target domain image which belongs to the same application scene with the target surface image;
continuing with the foregoing example, the same application scenario, that is, the corresponding detection application scenario in the silicon steel strip production mentioned above, is to capture and acquire an image (target area image) of the surface of the steel strip by using a specifically configured camera device, and it is easily understood that since the capture viewing angle, the occlusion condition, and the illumination condition are consistent, the obtained target area image has the same "style" (data distribution is consistent) as the image to be actually detected.
Then, step S120 is performed, the original source domain image with the label is processed according to the target domain image to obtain a source domain image with the data distribution characteristic of the target domain image, for example, in practice, a plurality of original source domain images can be respectively processed by using one target domain image to correspondingly obtain a plurality of source domain images with the data distribution characteristic of the target domain image;
it is easily understood that the original source domain image with labels here refers to the image in the training data set that people construct for training the model in other application scenarios. In this step, by performing specific processing, a source domain image having a target domain image data distribution characteristic (where the original annotation information is not affected) can be obtained, or the processed image has a "style" of the target domain image.
Then, step S130 is performed, and the source domain image having the target domain image data distribution characteristic in step S120 is used to perform supervised training on the model, so as to obtain a trained defect region segmentation model.
It is easy to understand that the source domain image with the target domain image data distribution characteristic is adopted to perform supervised training on the model, and the obtained trained model has better performance in the application scene corresponding to the target domain image, and can realize higher-precision defect region segmentation. By adopting the mode, manual marking of target domain data of practical application is not needed, and model training (supervised learning) is carried out based on marked data, so that the implementation mode of the application is higher in efficiency on the whole.
According to the technical scheme, in the model training process, the original source domain image with the label is processed through the target domain image to obtain the source domain image with the target domain image data distribution characteristic, the source domain image with the target domain image data distribution characteristic is used for carrying out supervised training on the model, and by adopting the model training mode, the performance of segmenting the model in various application scenes is improved more efficiently, so that the accuracy of segmenting the defect region of the image in specific application scenes is improved.
To facilitate understanding of the technical solutions of the present application, the technical solutions of the present application will be described below with reference to another embodiment.
In this embodiment, similarly to the previous embodiments, the following is also included:
acquiring a target surface image;
and secondly, performing defect region segmentation processing on the target surface image by using the trained defect region segmentation model so as to realize the segmentation of the surface defect region.
In this embodiment, the training process of the defect region segmentation model includes:
step 1, acquiring a target domain image belonging to the same application scene as a target surface image;
then, step 2 is carried out, the original source domain image with the label is processed according to the target domain image, and a source domain image with the data distribution characteristic of the target domain image is obtained;
specifically, in step 2, as shown in fig. 2, fourier transform is performed on the target domain image and the original source domain image, respectively, to obtain a first magnitude spectrum (corresponding to the source domain magnitude in fig. 2) and a first phase spectrum (corresponding to the source domain phase in fig. 2) corresponding to the original source domain image, and obtain a second magnitude spectrum (corresponding to the target domain magnitude in fig. 2) and a second phase spectrum (corresponding to the target domain phase in fig. 2) corresponding to the target domain image;
replacing low-frequency part data in the first amplitude spectrum with low-frequency part data in the second amplitude spectrum to obtain a third amplitude spectrum (not shown in fig. 2) after replacement, wherein the low-frequency part data is spectrogram data of which the frequency in the amplitude spectrum is less than or equal to a preset threshold frequency;
performing Fourier inverse transformation according to the third magnitude spectrum and the first phase spectrum, and taking an image obtained after inverse transformation as a source domain image with the data distribution characteristic of the target domain image;
in order to further improve the scene pertinence of the method and ensure the precision, corresponding threshold frequency values in different application scenes are aimed at, as a specific implementation mode, in the above process of step 2, aiming at the threshold frequency values, in the implementation of the method, the application scene can be determined first, and then the value of the preset threshold frequency is determined based on the determined application scene, but for the convenience of the implementation of the method, the corresponding relation between the application scene and the threshold frequency values can be configured in advance, and the determination of the value of the preset threshold frequency is implemented based on the corresponding relation.
As a specific implementation manner, for the value of the preset threshold frequency, the value of the preset threshold frequency can be one third of the upper limit value of the amplitude spectrum frequency range through configuration, and this manner can also obtain a better final effect and is beneficial to the applicability of the method in more scenes (such as scenes not related to the corresponding relationship).
As a specific implementation manner, in the above process in step 2, the low-frequency part data in the first amplitude spectrum is replaced with the low-frequency part data in the second amplitude spectrum to obtain a replaced third amplitude spectrum, which specifically is:
determining a circular region to be scratched on a second amplitude spectrum by taking the middle point of the amplitude spectrum as the circle center and the preset length as the radius, wherein the preset length is determined based on the preset threshold frequency;
and extracting the spectrogram corresponding to the circular area on the second magnitude spectrum, and correspondingly replacing the spectrogram into the first magnitude spectrum to obtain a third magnitude spectrum.
And then, performing step 3, and performing supervised training on the model by using the source domain image with the target domain image data distribution characteristics in the step 2 to obtain a trained defect region segmentation model.
In the embodiment, in the process of the source domain image with the target domain image data distribution characteristic, a scheme of performing low-frequency partial data replacement based on fourier transform is specifically adopted, and compared with a mode of generating a countermeasure network (GAN) to realize unsupervised machine learning in the related art, the method is simple and efficient to realize, and is beneficial to popularization and application of the whole technical scheme in an industrial field.
Fig. 3 is a schematic structural diagram of an unsupervised domain adapted surface defect region segmentation apparatus according to an embodiment of the present application, and as shown in fig. 3, the unsupervised domain adapted surface defect region segmentation apparatus 300 includes:
an obtaining module 301, configured to obtain a target surface image;
a segmentation processing module 302, configured to perform defect region segmentation processing on the target surface image by using the trained defect region segmentation model;
a training module 303, configured to train the defect region segmentation model, where the training module 303 is specifically configured to:
acquiring a target domain image belonging to the same application scene as the target surface image;
processing the original source domain image with the label according to the target domain image to obtain a source domain image with the data distribution characteristic of the target domain image;
and carrying out supervised training on the model by utilizing the source domain image with the target domain image data distribution characteristic.
Further, in the specific configuration process of the training module 303, processing the original source domain image with the label according to the target domain image to obtain a source domain image with the data distribution characteristic of the target domain image, including:
respectively carrying out Fourier transform on the target domain image and the original source domain image to obtain a first amplitude spectrum and a first phase spectrum corresponding to the original source domain image and obtain a second amplitude spectrum and a second phase spectrum corresponding to the target domain image;
replacing low-frequency part data in the first amplitude spectrum with low-frequency part data in the second amplitude spectrum to obtain a replaced third amplitude spectrum;
performing Fourier inverse transformation according to the third magnitude spectrum and the first phase spectrum, and taking an image obtained after inverse transformation as a source domain image with the data distribution characteristic of the target domain image;
the low-frequency part data is spectrogram data with the frequency less than or equal to a preset threshold frequency in the amplitude spectrum.
Further, in the specific configuration process of the training module 303, the low-frequency part data in the first amplitude spectrum is replaced with the low-frequency part data in the second amplitude spectrum to obtain a replaced third amplitude spectrum, which specifically includes:
determining a circular region to be scratched on a second amplitude spectrum by taking the middle point of the amplitude spectrum as the circle center and the preset length as the radius, wherein the preset length is determined based on the preset threshold frequency;
and extracting a spectrogram corresponding to the circular area on the second magnitude spectrum, and correspondingly replacing the spectrogram into the first magnitude spectrum to obtain a third magnitude spectrum.
Further, the training module 303 further comprises a preset threshold frequency determination sub-module (not shown in the figure) configured to:
determining an application scene; and determining the value of the preset threshold frequency based on the determined application scene.
In another specific embodiment, the preset threshold frequency determination sub-module is configured to,
and acquiring frequency range information of the amplitude spectrum, and taking one third of the upper limit value of the frequency range of the amplitude spectrum as a value of the preset threshold frequency.
With respect to the unsupervised domain adapted surface defect region segmentation apparatus 300 in the above-described related embodiment, the specific manner in which each module performs operations has been described in detail in the embodiment related to the method, and will not be elaborated herein.
Fig. 4 is a schematic structural diagram of an electronic device according to an embodiment of the present application, and as shown in fig. 4, the electronic device 400 includes:
a memory 401 having an executable program stored thereon;
a processor 402 for executing the executable program in the memory 401 to implement the steps of the above method.
With respect to the electronic device 400 in the above embodiment, the specific manner of executing the program in the memory 401 by the processor 402 thereof has been described in detail in the embodiment related to the method, and will not be elaborated herein.
The above description is only a preferred embodiment of the present invention, but the scope of the present invention is not limited thereto, and any changes or substitutions that can be easily conceived by those skilled in the art within the technical scope of the present invention are included in the scope of the present invention. Therefore, the protection scope of the present invention shall be subject to the protection scope of the claims.
Claims (8)
1. An unsupervised domain adapted surface defect region segmentation method, comprising:
acquiring a target surface image;
utilizing the trained defect region segmentation model to carry out defect region segmentation processing on the target surface image;
the training process of the defect region segmentation model comprises the following steps:
acquiring a target domain image belonging to the same application scene as the target surface image;
processing the original source domain image with the label according to the target domain image to obtain a source domain image with the data distribution characteristic of the target domain image;
and carrying out supervised training on the model by using the source domain image with the target domain image data distribution characteristics to obtain the trained defect region segmentation model.
2. The method according to claim 1, wherein the processing the original source domain image with the label according to the target domain image to obtain the source domain image with the target domain image data distribution characteristic comprises:
respectively carrying out Fourier transform on the target domain image and the original source domain image to obtain a first magnitude spectrum and a first phase spectrum corresponding to the original source domain image and obtain a second magnitude spectrum and a second phase spectrum corresponding to the target domain image;
replacing low-frequency part data in the first amplitude spectrum with low-frequency part data in the second amplitude spectrum to obtain a replaced third amplitude spectrum;
performing inverse Fourier transform according to the third magnitude spectrum and the first phase spectrum, and taking an image obtained after inverse transform as the source domain image with the target domain image data distribution characteristic;
the low-frequency part data is spectrogram data with the frequency less than or equal to a preset threshold frequency in the amplitude spectrum.
3. The method of claim 2, further comprising:
determining an application scene;
and determining the value of the preset threshold frequency based on the determined application scene.
4. The method of claim 2, wherein the predetermined threshold frequency is one third of the upper limit of the frequency range of the amplitude spectrum.
5. The method according to claim 2, wherein the low-frequency part data in the first amplitude spectrum is replaced by the low-frequency part data in the second amplitude spectrum to obtain a replaced third amplitude spectrum, and specifically:
determining a circular region to be scratched on a second amplitude spectrum by taking the middle point of the amplitude spectrum as the circle center and a preset length as the radius, wherein the preset length is determined based on the preset threshold frequency;
and extracting a spectrogram corresponding to the circular area on the second magnitude spectrum, and correspondingly replacing the spectrogram into the first magnitude spectrum to obtain the third magnitude spectrum.
6. The method of claim 1, wherein the defect region segmentation model is a PSPnet model or a U-Net model.
7. An unsupervised domain adapted surface defect region segmentation apparatus, comprising:
the acquisition module is used for acquiring a target surface image;
the segmentation processing module is used for carrying out defect region segmentation processing on the target surface image by utilizing the trained defect region segmentation model;
a training module, configured to train the defect region segmentation model, and specifically configured to:
acquiring a target domain image belonging to the same application scene as the target surface image;
processing the original source domain image with the label according to the target domain image to obtain a source domain image with the data distribution characteristic of the target domain image;
and carrying out supervised training on the model by utilizing the source domain image with the target domain image data distribution characteristic.
8. An electronic device, comprising:
a memory having an executable program stored thereon;
a processor for executing the executable program in the memory to implement the steps of the method of any one of claims 1-6.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111230003.0A CN113674277B (en) | 2021-10-22 | 2021-10-22 | Unsupervised domain adaptive surface defect region segmentation method and device and electronic equipment |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111230003.0A CN113674277B (en) | 2021-10-22 | 2021-10-22 | Unsupervised domain adaptive surface defect region segmentation method and device and electronic equipment |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN113674277A true CN113674277A (en) | 2021-11-19 |
| CN113674277B CN113674277B (en) | 2022-02-22 |
Family
ID=78550902
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202111230003.0A Active CN113674277B (en) | 2021-10-22 | 2021-10-22 | Unsupervised domain adaptive surface defect region segmentation method and device and electronic equipment |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113674277B (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114550215A (en) * | 2022-02-25 | 2022-05-27 | 北京拙河科技有限公司 | Target detection method and system based on transfer learning |
| CN115272318A (en) * | 2022-09-27 | 2022-11-01 | 精技精密部件(南通)有限公司 | Surface self-adaptive defect detection method for silicon steel strip |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108414226A (en) * | 2017-12-25 | 2018-08-17 | 哈尔滨理工大学 | Fault Diagnosis of Roller Bearings under the variable working condition of feature based transfer learning |
| US20190286073A1 (en) * | 2018-03-16 | 2019-09-19 | Salesforce.Com, Inc. | Systems and methods for learning for domain adaptation |
| CN112116102A (en) * | 2020-09-27 | 2020-12-22 | 张洪铭 | Method and system for expanding domain adaptive training set |
| CN112991353A (en) * | 2021-03-12 | 2021-06-18 | 北京航空航天大学 | Unsupervised semantic segmentation method for cross-domain remote sensing image |
| CN112990259A (en) * | 2021-02-04 | 2021-06-18 | 西交利物浦大学 | Early fault diagnosis method of rotary mechanical bearing based on improved transfer learning |
| CN113221902A (en) * | 2021-05-11 | 2021-08-06 | 中国科学院自动化研究所 | Cross-domain self-adaptive semantic segmentation method and system based on data distribution expansion |
-
2021
- 2021-10-22 CN CN202111230003.0A patent/CN113674277B/en active Active
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108414226A (en) * | 2017-12-25 | 2018-08-17 | 哈尔滨理工大学 | Fault Diagnosis of Roller Bearings under the variable working condition of feature based transfer learning |
| US20190286073A1 (en) * | 2018-03-16 | 2019-09-19 | Salesforce.Com, Inc. | Systems and methods for learning for domain adaptation |
| CN112116102A (en) * | 2020-09-27 | 2020-12-22 | 张洪铭 | Method and system for expanding domain adaptive training set |
| CN112990259A (en) * | 2021-02-04 | 2021-06-18 | 西交利物浦大学 | Early fault diagnosis method of rotary mechanical bearing based on improved transfer learning |
| CN112991353A (en) * | 2021-03-12 | 2021-06-18 | 北京航空航天大学 | Unsupervised semantic segmentation method for cross-domain remote sensing image |
| CN113221902A (en) * | 2021-05-11 | 2021-08-06 | 中国科学院自动化研究所 | Cross-domain self-adaptive semantic segmentation method and system based on data distribution expansion |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114550215A (en) * | 2022-02-25 | 2022-05-27 | 北京拙河科技有限公司 | Target detection method and system based on transfer learning |
| CN115272318A (en) * | 2022-09-27 | 2022-11-01 | 精技精密部件(南通)有限公司 | Surface self-adaptive defect detection method for silicon steel strip |
| CN115272318B (en) * | 2022-09-27 | 2022-12-23 | 精技精密部件(南通)有限公司 | Surface self-adaptive defect detection method for silicon steel strip |
Also Published As
| Publication number | Publication date |
|---|---|
| CN113674277B (en) | 2022-02-22 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108961217B (en) | Surface defect detection method based on regular training | |
| CN113674277B (en) | Unsupervised domain adaptive surface defect region segmentation method and device and electronic equipment | |
| CN110148130B (en) | Method and device for detecting part defects | |
| CN104902267B (en) | No-reference image quality evaluation method based on gradient information | |
| Nafchi et al. | Efficient no-reference quality assessment and classification model for contrast distorted images | |
| CN112837290B (en) | Crack image automatic identification method based on seed filling algorithm | |
| CN109472788B (en) | Method for detecting flaw on surface of airplane rivet | |
| CN111161222B (en) | Printing roller defect detection method based on visual saliency | |
| CN110378227B (en) | Method, device and equipment for correcting sample labeling data and storage medium | |
| RU2008129793A (en) | METHOD FOR IMPROVING FURTHER PROCESSING OF IMAGES USING DEFORMABLE NETS | |
| CN111339902B (en) | Liquid crystal display indication recognition method and device for digital display instrument | |
| CN108537787B (en) | Quality judgment method for face image | |
| CN112766110A (en) | Training method of object defect recognition model, object defect recognition method and device | |
| CN104182983B (en) | Highway monitoring video definition detection method based on corner features | |
| CN113284154B (en) | Steel coil end face image segmentation method and device and electronic equipment | |
| CN103413149A (en) | Method for detecting and identifying static target in complicated background | |
| CN117094975A (en) | Method and device for detecting surface defects of steel and electronic equipment | |
| Tian et al. | Corrosion identification of fittings based on computer vision | |
| Peng et al. | Non-uniform illumination image enhancement for surface damage detection of wind turbine blades | |
| CN105229665A (en) | To the enhancing analysis of the snakelike belt wear assessment based on image | |
| Yildirim et al. | Simultaneously realization of image enhancement techniques on real-time FPGA | |
| CN116596895A (en) | Substation equipment image defect identification method and system | |
| CN112183224A (en) | Model training method for image recognition, image recognition method and device | |
| CN111307267A (en) | Conductor galloping monitoring method based on concentric circle detection | |
| Danilescu et al. | Road anomalies detection using basic morphological algorithms |
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 |