[Paper Review] Fabric Defect Detection Using Vision-Based Tactile Sensor

Bin Fang, Xingming Long+4 more|arXiv (Cornell University)|Mar 2, 2020

Industrial Vision Systems and Defect Detection7 references4 citations

TL;DR

This paper proposes a vision-based tactile sensor system for fabric defect detection that captures surface texture without relying on color or pattern, using a camera, LEDs, and an elastic sensing layer. By employing intensity adjustment, ResNet with ensemble learning, and uniformity-based dataset selection, the system achieves robust detection of structural defects across diverse fabrics without external lighting, demonstrating strong feasibility and performance in real-world conditions.

ABSTRACT

This paper introduces a new type of system for fabric defect detection with the tactile inspection system. Different from existed visual inspection systems, the proposed system implements a vision-based tactile sensor. The tactile sensor, which mainly consists of a camera, four LEDs, and an elastic sensing layer, captures detailed information about fabric surface structure and ignores the color and pattern. Thus, the ambiguity between a defect and image background related to fabric color and pattern is avoided. To utilize the tactile sensor for fabric inspection, we employ intensity adjustment for image preprocessing, Residual Network with ensemble learning for detecting defects, and uniformity measurement for selecting ideal dataset for model training. An experiment is conducted to verify the performance of the proposed tactile system. The experimental results have demonstrated the feasibility of the proposed system, which performs well in detecting structural defects for various types of fabrics. In addition, the system does not require external light sources, which skips the process of setting up and tuning a lighting environment.

Motivation & Objective

• To address the challenge of false positives in visual fabric inspection caused by color and pattern variations.
• To develop a tactile sensing system that captures surface texture independently of visual appearance.
• To eliminate the need for controlled lighting environments in fabric defect detection systems.
• To improve defect detection accuracy across diverse fabric types using deep learning and dataset optimization.
• To create a self-contained, lighting-independent inspection system suitable for industrial deployment.

Proposed method

• The system uses a vision-based tactile sensor with a camera, four LEDs, and an elastic sensing layer to capture surface microstructures.
• Image preprocessing applies intensity adjustment to normalize input for consistent defect visibility.
• A Residual Network (ResNet) with ensemble learning is used to detect structural defects in fabric images.
• Uniformity measurement is applied to select high-quality, representative training samples for model generalization.
• The system operates without external light sources, relying on internal LED illumination for consistent imaging.
• The entire pipeline is designed to minimize false positives caused by fabric color and pattern variations.

Experimental results

Research questions

• RQ1 Can a vision-based tactile sensor effectively detect fabric defects without being influenced by color and pattern variations?
• RQ2 How does the proposed system perform in detecting structural defects across diverse fabric types?
• RQ3 To what extent can intensity adjustment and uniformity-based dataset selection improve defect detection accuracy?
• RQ4 Can the system operate reliably without external lighting, simplifying industrial integration?
• RQ5 What is the feasibility and robustness of the proposed tactile sensing approach in real-world fabric inspection?

Key findings

• The system successfully detects structural defects in various fabric types without being affected by color or pattern variations.
• The absence of external lighting requirements simplifies system setup and enhances deployment flexibility in industrial environments.
• Intensity adjustment and uniformity-based dataset selection significantly improve model training stability and generalization.
• The use of ResNet with ensemble learning achieves high accuracy in identifying subtle surface defects.
• Experimental results confirm the feasibility and robustness of the proposed tactile sensing approach for fabric inspection.
• The system demonstrates strong performance across diverse fabric types, indicating broad applicability.