Automated Defect Detection and Characterization on Pulse Thermography Images Using Computer Vision Techniques

Abstract

Defect detection and characterization plays a vital role in predicting the life span of materials. Defect detection using appropriate inspection technologies at various phases has gained huge importance in metal production lines. It can be accomplished through wise application of non-destructive testing and evaluation (NDE). It is important to characterize defects at an early stage in order to be able to overcome them or take corrective measures. Pulse thermography is a modern NDE method that can be used for defect detection in metal objects. Only a limited amount of work has been done on automated detection and characterization of defects due to thermal diffusion. This paper proposes a system for automatic defect detection and characterization in metal objects using pulse thermography images as well as various image processing algorithms and mathematical tools. An experiment was carried out using a sequence of 250 pulse thermography images of an AISI 316 L stainless steel sheet with synthetic defects. The proposed system was able to detect and characterize defects sized 10 mm, 8 mm, 6 mm, 4 mm and 2 mm with an average accuracy of 96%, 95%, 84%, 77%, 54% respectively. The proposed technique helps in the effective and efficient characterization of defects in metal objects.