RPDNet: Automatic Fabric Defect Detection Based on a Convolutional Neural Network and Repeated Pattern Analysis.

Area of Science:

• Textile manufacturing and quality control
• Computer vision and machine learning
• Image processing for defect detection

Background:

• Automatic defect detection is crucial for textile quality control.
• Existing methods may lack efficiency or require significant computational resources.
• Identifying periodic patterns in fabric is key to accurate defect localization.

Purpose of the Study:

• To propose a novel semantic segmentation network, RPDNet, for pixel-level fabric defect detection.
• To enhance defect detection accuracy by utilizing repeated pattern analysis.
• To develop a lightweight and efficient model for real-time applications.

Main Methods:

• Developed a repeated pattern detector using Convolutional Neural Networks (CNNs).
• Integrated a semi-supervised learning scheme to inject periodic knowledge into the model.
• Utilized DeeplabV3+ and GhostNet architectures for a lightweight design.

Main Results:

• RPDNet achieves competitive detection results on repeated pattern fabric images.
• The model effectively identifies potential defect areas by understanding periodic features.
• The semi-supervised approach allows independent model function without pre-calculation, maintaining detection speed.

Conclusions:

• RPDNet offers an efficient and accurate solution for automatic fabric defect detection.
• The integration of repeated pattern analysis and advanced network architectures leads to superior performance.
• The proposed method provides a cost-effective approach to textile quality control.