.png){kind=icon}
.png){kind=icon}
Return
Fabric defects detection method based on siamese network model
.png){kind=icon}
PRE
AI
DOI：10.1080/00405000.2025.2607851.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
Fabric defect is an accidental phenomenon in the process of fabric production, and there are many kinds of fabric defects with different shapes. In order to improve the detection efficiency, this paper proposes a phased fabric defect detection method based on siamese network model. By analyzing the similarity detection mechanism of siamese network, a similarity detection model based on siamese network is constructed. The model consists of three modules: feature extraction module, similarity measurement module and defect classification module. Firstly, Resnet is introduced to build a feature extraction module to extract multi-scale features of sample pairs. Secondly, in the similarity measurement module, the CA attention mechanism is introduced to improve the feature difference between the defect-free samples and the defect samples. The abstract features of the samples are transformed into similarity calculation. The sensitivity of the Siamese network to the difference is used to realize the rapid discrimination of whether the fabric image contains defects, that is, the task of the defect discrimination stage is completed. Finally, in the defect classification module, the defect image is input into the Yolov4 network to achieve accurate defect classification, that is, to complete the task of defect classification stage. The experimental results show that the proposed method can effectively detect fabric defects, and the detection accuracy reaches 87.54%, the defect-free detection speed reaches 126.58 fps, and the defect detection speed reaches 62.84 fps, which provides a new method for fabric defect detection.
Keywords:
Fabric defect
object detection
siamese network
Yolo
Journal
J
IF:
1.5
Papers:
116
Citations:
5.4K