Study on Crack Features in Images of Fluorescent Magnetic Particle Inspection for Railway Wheelsets

Tao Ma,Zhenguo Sun,Qiang Chen
DOI: https://doi.org/10.1784/insi.2018.60.9.519
2018-01-01
Insight - Non-Destructive Testing and Condition Monitoring
Abstract:Crack features in the images of fluorescent magnetic particle inspection (FMPI for railway wheelsets are studied in this paper. Firstly, according to the curvilinear structure of the cracks, a ridge-line detection (RLD) approach is used to extract ridge lines of the image structures as suspected cracks. Secondly, the shapes of the ridge lines and their image structures are exploited, which leads to shape features being extracted from the ridge lines and an additional restriction in ridge-line detection. Thirdly, the local textures of the suspected cracks are analysed and a texture feature is proposed based on averaging the modified scale-invariant feature transform (SIFT) descriptors of the points on the suspected crack. Experiments on real FMPI images of railway wheelsets prove the effectiveness of the proposed features. Among the tested algorithms, the proposed algorithm, which uses the modified ridge-line detection approach to extract suspected cracks and the texture feature to distinguish real cracks from all suspected cracks, achieves the highest recall and precision, both of which are over 90%.
What problem does this paper attempt to address?