Application Of PMP In Online Measurement Of Rail Surface Defects

With the development of the technology, nondestructive testing technology especially three-dimensional nondestructive testing have received the widespread attention and research. Phase Measurement Profilometry (PMP), as a kind of optical three-dimensional contour Measurement detection, is widely used in industrial inspection, medical diagnostics, physical shape, and other fields because of the advantages its non-contact, non-destructive, high accuracy. As an important link in national transportation, railway transportation carries most of the passenger and freight transport, thus the railway safety detection has received the widespread attention. With the growing demand of social development, railway is developing towards high-speed, overloading and high density, and Rail as an important part of the train operation, the safety test should not be ignored. The technique of phase measuring profilometry is used to detect rail surface defect online.In this thesis, the situation of the rail wear and rail surface defects detection at home and abroad is studied. After that the phase measuring profilometry is researched and analyzed. By the comparison and analysis of existing phase shift algorithm, the traditional Stoilov phase shift algorithm is studied, and an improved Stoilov algorithm based on probability and statistics is proposed. Improved Stoilov algorithm can effectively remove the four kinds of singular points in the traditional algorithm and ensure the precision of phase measuring profilometry.Pixel matching is a vital part of online phase measuring profilometry. In this thesis, the existing pixel matching algorithms and image matching methods are investigated and researched, then the SIFT algorithm is introduced into the pixel matching. To measure the 3D profile of the rail, the new sub-pixel matching algorithm based on SIFT and image grayscale is put forward. This algorithm can achieve sub-pixel level matching, and does not require special markers on conveyor belt. This method is invariant to the shape and the size of the object to be tested, and can guarantee the precision of phase measuring profilometry online.Through experimental verification and error analysis, using the improved phase shift algorithm and pixel matching algorithm, can satisfy the requirement of online rail three-dimensional contour detection. The experimental results show that the phase measuring profilometry is applied to rail on-line detection is feasible.