Evaluation of anomaly detection capabilities using a non-orthogonal camera angle in pulse-phase thermography

Abstract Pulse-phase thermography (PPT) is widely used to nondestructively inspect internal defects in fiber reinforced polymers. However, the challenges using PPT for complex shapes is poorly documented in literature. Only small changes in the object distance have been considered. Complex parts can have significant variations in object distance and thus, in detected radiation. In this contribution, the effect of a non-orthogonal camera angle with respect to a flat sample, leading to varying object distances and an inhomogeneous sound background area in phasegrams, is investigated. Samples with artificial round and square defects of different sizes are positioned under varying angles with respect to the camera, representing geometric properties of complex parts. The construction of the thermographic system and the experimental setup to systematically vary the angle between camera and specimen is presented. We investigated the change of the signal-to-noise ratio (SNR) of artificial delaminations in PPT measurements under varying object distances. The SNR in a distance of 136 mm out of the focal plane is sufficiently high for image feature extraction. Phasegrams are exported to a colored representation, leading to a higher contrast in distinct color channels. An algorithm which extracts and merges defect information from three different color channels is developed. Challenging lighting conditions lead to a noisy background having artifacts. The developed filter performs better in defect detection and size quantification than a global or local threshold in grayscale phasegrams under those conditions.