Automatic Defect Detection in Translucent Materials Using Digital Photoelasticity and GLCM-Based Homogeneity Analysis

Abstract

Transparent and translucent polymeric materials, such as polycarbonate, are widely used in industrial applications due to their mechanical strength, low weight, and optical properties. Nevertheless, internal defects and stress concentrations are often difficult to detect by conventional visual inspection methods, requiring the use of alternative non-destructive evaluation techniques. The objective of this work is to develop a lightweight computer vision system capable of qualitatively detecting structural defects in translucent materials using digital photoelasticity and texture analysis. The proposed methodology employs a digital plane polariscope comprising a whitelight source, linear polarizers, and a digital imaging sensor, operating under controlled laboratory conditions. The images were processed through grayscale conversion, preprocessing stages, and Gray-Level Co-occurrence Matrix (GLCM)-based homogeneity analysis to generate spatial heatmaps highlighting texture perturbations associated with defects. The experimental tests were performed on healthy and intentionally damaged polycarbonate samples with various geometries. The obtained results demonstrate that the proposed system can qualitatively localize anomalous regions without requiring computationally intensive artificial intelligence models. These findings validate the feasibility of combining digital photoelasticity and lightweight computer vision techniques for non-destructive inspection and industrial quality control applications.