Enhanced Metal Surface Defect Localization with SH-SAM and Grey Level Co-Occurrence Matrix (GLCM) in Robotics Automation

Abstract

The identification of metal surface defects is essential for many sectors, including aerospace and automotive, in the production of quality products. Traditional approaches are time-consuming and error-prone and, hence, an automated, accurate fault localisation solution in robotic automation is necessary. This research aims to enhance the defect detection process by integrating SH-SAM with GLCM for enhanced precision and robustness in robotic automation. We therefore aim at developing a whole system that integrates texture with spectral analysis for accurate identification of defects. The proposed method integrates Supervised Hyperspectral Anomaly Detection (SH-SAM) with a grey-level co-occurrence Matrix (GLCM) to analyze spectral anomalies and texture patterns on metal surfaces. This integration enhances the discovery of flaws by robotic automation. The approach that combines SH-SAM and GLCM performed better compared to any of the methods individually on F1 score (88.5%), accuracy (92%), and precision (89%). It also outperformed all the others in defect localization as RME was decreased to 9.3%. The integration of SH-SAM and GLCM offers a highly effective solution for the localisation of metal surface defects with improved accuracy and reduced errors. This method shows great potential for real-time robotic automation in metal surface inspection applications.