Image time series vectorization for material debond characterization using laser spot thermography

Abstract

In the aircraft, civil and related industries is important to check the quality and health of the key components during the manufacturing process or the maintenance examinations. The most common found problems are the debonds present at the material surface. The laser spot thermography is a promising technology for the localization of the surface cracks. For this purpose, we built an experimental setup for laser spot control over the sample surface. Using an infrared (IR) thermal camera, we registered image time series of various experimental samples irradiated with the controlled laser spot. The concentrated laser beam passing over the material sample cracks causing a lateral heat flux disrupted by the presence of surface breaking fractures of different widths. The image time series obtained using measurements and simulations were used to generate input features for a neural network classifier which is trained for the determination of the crack width. Image processing workflow for features generation includes image enhancement, anisotropic diffusion as preprocessing techniques, crack detection using cellular neural networks CNN and also the description of the detected crack signature. Crack signature descriptions aim for the qualification of the sample under analysis for further use in various applications were nondestructive material debond characterization is needed.