CN115984215A - Fiber bundle defect detection method based on twin network - Google Patents
Fiber bundle defect detection method based on twin network Download PDFInfo
- Publication number
- CN115984215A CN115984215A CN202211711576.XA CN202211711576A CN115984215A CN 115984215 A CN115984215 A CN 115984215A CN 202211711576 A CN202211711576 A CN 202211711576A CN 115984215 A CN115984215 A CN 115984215A
- Authority
- CN
- China
- Prior art keywords
- fiber bundle
- defect
- image
- twin network
- defect detection
- 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
- 239000000835 fiber Substances 0.000 title claims abstract description 107
- 230000007547 defect Effects 0.000 title claims abstract description 86
- 238000001514 detection method Methods 0.000 title claims abstract description 57
- 238000012549 training Methods 0.000 claims abstract description 11
- 230000002950 deficient Effects 0.000 claims description 30
- 239000013598 vector Substances 0.000 claims description 10
- 238000013528 artificial neural network Methods 0.000 claims description 6
- 238000000605 extraction Methods 0.000 claims description 3
- 238000000034 method Methods 0.000 abstract description 7
- 238000002372 labelling Methods 0.000 abstract description 2
- 230000006870 function Effects 0.000 description 8
- 238000010586 diagram Methods 0.000 description 4
- 238000003860 storage Methods 0.000 description 4
- 238000013135 deep learning Methods 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 238000004519 manufacturing process Methods 0.000 description 2
- 230000003287 optical effect Effects 0.000 description 2
- 229920000049 Carbon (fiber) Polymers 0.000 description 1
- 239000004917 carbon fiber Substances 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000009776 industrial production Methods 0.000 description 1
- 238000007689 inspection Methods 0.000 description 1
- VNWKTOKETHGBQD-UHFFFAOYSA-N methane Chemical compound C VNWKTOKETHGBQD-UHFFFAOYSA-N 0.000 description 1
- 238000012544 monitoring process Methods 0.000 description 1
- 238000012545 processing Methods 0.000 description 1
- 238000012797 qualification Methods 0.000 description 1
- 238000003908 quality control method Methods 0.000 description 1
- 239000002699 waste material Substances 0.000 description 1
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 discloses a fiber bundle defect detection method based on a twin network, which comprises the steps of obtaining a normal fiber bundle image and a defect fiber bundle image, inputting the normal fiber bundle image and the defect fiber bundle image into the twin network for training to obtain a defect detection model; and inputting the fiber bundle image to be detected into a trained twin network detection model for fiber bundle defect detection, and judging whether the fiber bundle image contains defects. According to the method, a reliable fiber bundle defect detection model can be obtained without acquiring a large number of defect fiber bundle images and only a small number of defect fiber bundle images, the problems that data are difficult to obtain in a real scene, and sample labeling is time-consuming and labor-consuming are solved, and the method has objective practical significance.
Description
Technical Field
The invention relates to the field of image processing, in particular to a fiber bundle defect detection method based on a twin network.
Background
Quality control is a crucial part in industrial production, and compared with quality spot check after production is completed, the requirement of high quality and low defect can be better met by performing online quality monitoring in the production process. With the development of computer vision and deep learning techniques, features can be extracted from images for surface defect detection, and the accuracy rate close to that of manual inspection is achieved.
At present, a computer vision technology based on a neural network is sensitive to data, a large amount of manpower and time are needed for model training and deployment in the industrial field, data acquisition in the industrial subdivision field is difficult, and application of deep learning in the industrial field is limited.
Disclosure of Invention
The invention provides a fiber bundle defect detection method based on a twin network, which adopts the twin neural network, can obtain a reliable fiber bundle defect detection model only by taking a small number of defective fiber bundle images as training samples, solves the problem of difficulty in obtaining defective samples in the industrial field, and has practical significance.
Aiming at the defects of the prior art, the invention discloses a fiber bundle defect detection method based on a twin network, which comprises the following steps:
step 1, acquiring a defect-free fiber bundle image and a defect fiber bundle image for manual marking;
step 2, inputting the non-defective fiber bundle image and the defective fiber bundle image into a twin network defect detection model, and training the twin network defect detection model to obtain a twin network defect detection model;
and 3, inputting the fiber bundle image to be detected into the trained twin network detection model for fiber bundle defect detection, and judging whether the fiber bundle image contains defects.
Further, the step 1 comprises: the image of the defect-free fiber bundle is marked as 0, and the image of the defect fiber bundle is marked as 1;
further, the step 2 comprises:
2-1, randomly selecting two images from the non-defective fiber bundle image and the defective fiber bundle image as a sample pair to be trained;
2-2, inputting the sample to be trained into the twin network for feature extraction, and respectively calculating feature vectors; the twin network is two VGG16 neural networks with the same structure and shared weight;
and 2-3, training the twin network defect detection model according to the loss function to obtain the twin network defect detection model.
Further, the loss function is defined as:
wherein L is i For the loss function, i is the number of the sample pair, (x) 1 ,x 2 T) is the pair of samples and the label, if x 1 ,x 2 If the fiber bundle image is not defective or defective, t =1, otherwise t =0,b is constant, max is a function of the maximum value, B i Is the Manhattan distance, B, of the feature vector i Is defined as:
B i ＝|v(x 1 )-v(x 2 )| i
wherein, v (x) 1 ) And v (x) 2 ) Is a feature vector.
Further, the step 3 comprises:
step 3-1, collecting a fiber bundle image to be detected;
step 3-2, inputting the non-defective fiber bundle image and the fiber bundle image to be detected into the trained twin network defect detection model to obtain a non-defective similarity index; the similarity index is the reciprocal of the Manhattan distance;
step 3-3, inputting the defect fiber bundle image and the fiber bundle image to be detected into the trained twin network defect detection model to obtain a defect similarity index;
3-4, if the defect-free similarity index is larger than or equal to the defect similarity index, determining the fiber bundle image to be detected as a qualified fiber bundle; and if the defect-free similarity index is smaller than the defect similarity index, the fiber bundle image to be detected is a defect fiber bundle.
Has the advantages that: the invention discloses a fiber bundle defect detection method based on a twin network, which is a specific application of computer vision and neural network technology in the industrial field, uses a twin network defect detection model, can train out a reliable and effective detection model without collecting a large number of samples, can achieve online surface defect detection through vision on the premise of not contacting products, realizes quality detection automation, and can be popularized and applied in a large range in industries with similar requirements.
Drawings
FIG. 1 is a schematic workflow diagram of a fiber bundle defect detection method based on a twin network according to an embodiment of the present invention;
FIG. 2 is a schematic diagram of the defect detection effect of the defect-free fiber bundle image and the fiber bundle image to be detected according to the embodiment of the invention.
FIG. 3 is a schematic diagram of a defect detection effect of a part of a defective fiber bundle image and a fiber bundle image to be detected according to an embodiment of the present invention.
Detailed Description
In order to make the aforementioned objects, features and advantages of the present invention comprehensible, embodiments accompanied with figures are described in further detail below.
The embodiment of the invention discloses a fiber bundle defect detection method based on a twin network, which is applied to the detection of fiber bundle surface defects in the carbon fiber industry, only needs to acquire a small number of fiber bundle images for model training and deployment, and can meet the accurate and rapid actual use requirements.
The invention discloses a fiber bundle defect detection method based on a twin network, a schematic diagram of a working flow is shown in figure 1, and the method comprises the following steps:
step 1, acquiring a non-defective fiber bundle image and a defective fiber bundle image for manual marking; the image of the defect-free fiber bundle is marked as 0, and the image of the defect fiber bundle is marked as 1;
step 2, inputting the image of the defect-free fiber bundle and the image of the defect fiber bundle into a twin network defect detection model, and training the twin network defect detection model to obtain a twin network defect detection model;
and 3, inputting the fiber bundle image to be detected into the trained twin network detection model for fiber bundle defect detection, and judging whether the fiber bundle image contains defects.
In one implementation, the step 2 includes:
2-1, selecting two images from the non-defective fiber bundle image and the defective fiber bundle image to form a sample pair to be trained;
2-2, inputting the sample to be trained into the twin network for feature extraction, and respectively calculating feature vectors; the twin network is two VGG16 neural networks with the same structure and shared weight;
and 2-3, training the twin network defect detection model according to the loss function to obtain the twin network defect detection model. The loss function is defined as:
wherein L is i For the loss function, i is the number of the sample pair, (x) 1 ,x 2 T) is the sample pair and label, if x 1 ,x 2 If the fiber bundle image is not defective or defective, t =1, otherwise t =0, B =2, max is a function of the maximum value, B i Manhattan distance, B, of the feature vector i Is defined as:
B i ＝|v(x 1 )-v(x 2 )| i
wherein, v (x) 1 ) And v (x) 2 ) Is a feature vector.
In one implementation, the step 3 includes:
step 3-1, collecting a fiber bundle image to be detected;
step 3-2, inputting one non-defective fiber bundle image and the fiber bundle image to be detected into the trained twin network defect detection model to obtain a non-defective similarity index; the similarity index is the reciprocal of the Manhattan distance Bi of the feature vector;
step 3-3, inputting the defect fiber bundle image and the fiber bundle image to be detected into the trained twin network defect detection model to obtain a defect similarity index;
3-4, if the defect-free similarity index is larger than or equal to the defect similarity index, determining the fiber bundle image to be detected as a qualified fiber bundle; and if the defect-free similarity index is smaller than the defect similarity index, the fiber bundle image to be detected is a defect fiber bundle.
As shown in fig. 2, the image of the defect-free fiber bundle and the image of the fiber bundle to be detected are input into the trained twinning network defect detection model, so as to obtain a similarity index of 0.018 of the defect-free fiber bundle; as shown in fig. 3, the defect fiber bundle image and the fiber bundle image to be detected are input into the trained twinned network defect detection model, and the defect similarity index is 3.564. And if the defect-free similarity index is smaller than the defect similarity index, the fiber bundle image to be detected is a defect fiber bundle.
The invention discloses a fiber bundle defect detection method based on a twin network, which can obtain a reliable fiber bundle defect detection model by only carrying out model training on a small number of normal and defective fiber bundle images without acquiring a large number of defective fiber bundle images, solves the problems of difficult data acquisition and time and labor waste of sample labeling in the industrial field, can be implemented quickly and effectively, reduces the defect rate, improves the qualification rate and further improves the product quality.
In a specific implementation, the present invention further provides a computer storage medium, where the computer storage medium may store a program, and when the program is executed, the program may include some or all of the steps in each embodiment of the dual-feature fused semi-global stereo matching method provided by the present invention. The storage medium may be a magnetic disk, an optical disk, a read-only memory (ROM), a Random Access Memory (RAM), or the like.
Those skilled in the art will readily appreciate that the techniques of the embodiments of the present invention may be implemented as software plus a required general purpose hardware platform. Based on such understanding, the technical solutions in the embodiments of the present invention may be essentially or partially implemented in the form of a software product, which may be stored in a storage medium, such as ROM/RAM, magnetic disk, optical disk, etc., and includes several instructions for enabling a computer device (which may be a personal computer, a server, or a network device, etc.) to execute the method according to the embodiments or some parts of the embodiments.
The same and similar parts in the various embodiments in this specification may be referred to each other. The above-described embodiments of the present invention should not be construed as limiting the scope of the present invention.
Claims (5)
1. A fiber bundle defect detection method based on a twin network is characterized by comprising the following steps:
step 1, acquiring a non-defective fiber bundle image and a defective fiber bundle image for manual marking;
step 2, inputting the image of the nondefective fiber bundle and the image of the defective fiber bundle into an twin network, and training the twin network to obtain a twin network defect detection model;
and 3, inputting the fiber bundle image to be detected into the trained twin network detection model for fiber bundle defect detection, and judging whether the fiber bundle image contains defects.
2. The twin network based fiber bundle defect detection method according to claim 1, wherein the step 1 comprises: the image of the defect-free fiber bundle is labeled as 0, and the image of the defect fiber bundle is labeled as 1.
3. The fiber bundle defect detection method based on the twin network according to claim 1, wherein the step 2 comprises:
2-1, randomly selecting two images from the non-defective fiber bundle image and the defective fiber bundle image as a sample pair to be trained;
2-2, inputting the sample to be trained into the twin network for feature extraction, and respectively calculating feature vectors; the twin network is two VGG16 neural networks with the same structure and shared weight;
and 2-3, training the twin network defect detection model according to the loss function to obtain the twin network defect detection model.
4. The twin network based fiber bundle defect detection method according to claim 1, wherein the steps 2-3 comprise: the loss function is defined as:
wherein L is i For the loss function, i is the number of the sample pair, (x) 1 ,x 2 T) is the sample pair and label, if x 1 ,x 2 If the fiber bundle image is not defective or defective, t =1, otherwise t =0,b is constant, max is a function of the maximum value, B i Manhattan distance, B, of the feature vector i Is defined as:
B i ＝|v(x 1 )-v(x 2 )| i
wherein, v (x) 1 ) And v (x) 2 ) Is a feature vector.
5. The fiber bundle defect detection method based on the twin network according to claim 1, wherein the step 3 comprises:
step 3-1, collecting a fiber bundle image to be detected;
step 3-2, inputting the non-defective fiber bundle image and the fiber bundle image to be detected into the trained twin network defect detection model to obtain a non-defective similarity index; the similarity index is the reciprocal of the Manhattan distance;
step 3-3, inputting the defect fiber bundle image and the fiber bundle image to be detected into the trained twin network defect detection model to obtain a defect similarity index;
3-4, if the defect-free similarity index is greater than or equal to the defect similarity index, determining that the fiber bundle image to be detected is a qualified fiber bundle; and if the defect-free similarity index is smaller than the defect similarity index, the fiber bundle image to be detected is a defect fiber bundle.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211711576.XA CN115984215A (en) | 2022-12-29 | 2022-12-29 | Fiber bundle defect detection method based on twin network |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211711576.XA CN115984215A (en) | 2022-12-29 | 2022-12-29 | Fiber bundle defect detection method based on twin network |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN115984215A true CN115984215A (en) | 2023-04-18 |
Family
ID=85962246
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202211711576.XA Pending CN115984215A (en) | 2022-12-29 | 2022-12-29 | Fiber bundle defect detection method based on twin network |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115984215A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116958739A (en) * | 2023-06-25 | 2023-10-27 | 南京矩视科技有限公司 | Attention mechanism-based carbon fiber channel real-time dynamic numbering method |
-
2022
- 2022-12-29 CN CN202211711576.XA patent/CN115984215A/en active Pending
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116958739A (en) * | 2023-06-25 | 2023-10-27 | 南京矩视科技有限公司 | Attention mechanism-based carbon fiber channel real-time dynamic numbering method |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111179251B (en) | Defect detection system and method using template comparison based on Siamese neural network | |
| CN108074231B (en) | A method for surface defect detection of magnetic sheet based on convolutional neural network | |
| CN109919934B (en) | A liquid crystal panel defect detection method based on multi-source domain deep transfer learning | |
| CN111080622B (en) | Neural network training method, workpiece surface defect classification and detection method and device | |
| CN107564002A (en) | Plastic tube detection method of surface flaw, system and computer-readable recording medium | |
| CN117392042A (en) | Defect detection method, defect detection apparatus, and storage medium | |
| CN113487533B (en) | A digital inspection system and method for parts assembly quality based on machine learning | |
| CN111860596B (en) | Unsupervised road crack classification method and model establishment method based on deep learning | |
| CN113083804A (en) | Laser intelligent derusting method and system and readable medium | |
| CN114037645A (en) | Coating defect detection method, device, electronic device and readable medium for pole piece | |
| Liao et al. | Solder joint defect inspection method based on ConvNeXt-YOLOX | |
| CN112070727A (en) | Metal surface defect detection method based on machine learning | |
| CN113591948A (en) | Defect pattern recognition method and device, electronic equipment and storage medium | |
| CN111311545A (en) | Container detection method, device and computer readable storage medium | |
| CN113506239A (en) | Strip steel surface defect detection method based on cross-stage local network | |
| CN116152498A (en) | Data-driven Semantic Segmentation Network and Training Method for Metal Surface Defects | |
| CN115457026A (en) | Paper defect detection method based on improved YOLOv5 | |
| CN114429461A (en) | A cross-scenario strip surface defect detection method based on domain adaptation | |
| CN115984215A (en) | Fiber bundle defect detection method based on twin network | |
| CN114612468B (en) | Equipment external defect detection method based on positive sample | |
| CN111768380A (en) | Method for detecting surface defects of industrial spare and accessory parts | |
| CN115294039A (en) | Steel coil end surface defect detection method | |
| CN112668365A (en) | Material warehousing identification method, device, equipment and storage medium | |
| CN117911796B (en) | Intelligent data processing system and method based on image recognition | |
| CN118052809B (en) | Zero sample point cloud anomaly detection method and system considering prompt learning |
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 |