Repeated pattern detection on fabric: A survey and novel approach

Area of Science:

• Computer Vision
• Textile Manufacturing
• Pattern Recognition

Background:

• Textile industry uses repeating patterns for aesthetics.
• Computer vision is crucial for automating production and reducing waste.
• Detecting patterns on fabrics presents challenges like variable lighting and pattern variance.

Purpose of the Study:

• To develop a novel method for automatic detection of repeating patterns on fabric images.
• To achieve sub-pixel accuracy in pattern detection.
• To address real-world complexities in fabric pattern analysis.

Main Methods:

• Conducted an in-depth literature review on repeated pattern detection.
• Proposed a novel method for detecting patterns with sub-pixel accuracy.
• Performed extensive experiments comparing the proposed method against baselines.
• Utilized an automatic calibration procedure with limited human supervision, eliminating the need for training data.

Main Results:

• Achieved a 96% recall rate on a synthetic dataset.
• Demonstrated an average alignment error of less than 0.5 pixels.
• The method operates efficiently, providing results in seconds.
• Showcased high precision in real-world applications without requiring training data.

Conclusions:

• The proposed method offers a competitive and accurate solution for repeating pattern detection in textiles.
• It effectively handles real-world complexities, making it suitable for industrial automation.
• The release of the dataset and code aims to foster further research in this domain.