Edge Detection Algorithm When Applied To Pulse Thermographic Images

Abstract

Nondestructive testing schemes are widely used in several industries to assure quality. Such schemes include a range of testing techniques with several embodiments, such as ultrasound, penetrant testing and infrared thermography. This study focuses on the application of Infrared Thermography (IRT) and its direct implementation in carbon composites inspection. Thus, the coverage and scope of presented text addresses an active thermography setup with a pulsed stimulation approach equipped with a machine vision type processing sequences. The presented study investigates the edge detection operators and kernels effect on the results of the thermographic implementations to the composite specimens. Such analysis includes several known operators such as Sobel kernel, in addition to varying sequences of erosion and dilation, and furthermore a selfreferencing thresholding scheme. The image processing will address the morphological aspects of the defects' detection in addition to the effect of depth and heat diffusion. To this end, a test campaign is implemented on a set of test CFRP coupons. The results highlight the need to adjust the current machine vision edge detectors for thermographic applications while showing an improvement in detectability when self- referencing is used as a thresholding approach.

Date of Award	Dec 2016
Original language	American English
Supervisor	Mohammad Omar (Supervisor)