Efficient data evaluation for thermographic crack detection

We present an all-purpose crack detection algorithm for flying spot thermography which is directly applicable to a thermogram sequence without the need of any additional information about the experimental setup. A single image containing distinct crack signatures is derived in two steps. Firstly, the spatial derivative is calculated for each frame of the sequence and, secondly, the resulting data set is sorted pixel wise along the time axis. The feasibility of the proposed procedure is proven by testing a piece of rail that comprises roll contact fatigue cracks and by comparing the results with magnetic particle testing.