Terahertz non-destructive inner damages characterisation of GFRP composites using enhanced Canny edge detector

Terahertz (THz)-based techniques have proven to be effective non-destructive testing (NDT) approaches for investigating non-metallic and non-polar materials owing to their excellent penetration ability and transverse resolution. When investigating glass fibre reinforced polymer (GFRP) composites along the z-axis, THz images often suffer from low resolution and poor contrast in damage edge areas. This is primarily due to limitations in the power level and detector sensitivity of THz equipment, as well as unexpected noise interference during measurements. In this paper, an enhanced Canny edge detector is proposed to provide a clear representation of damage buried in GFRP composites. Compared with traditional Canny operator, Bilateral filter and Otsu segmentation algorithm are adopted to smooth THz images and determine the optimal two threshold values automatically. Experimental results show that the transverse resolution of THz images as well as damage contours improved significantly after enhanced Canny operator, and better three-dimensional rendering performance is achieved, demonstrating that THz time-of-flight tomography combined with proposed computer vision algorithms is capable of accurately locating and clearly visualising damage in composite materials in a non-destructive and non-contact fashion.