Key Techniques Of 3D Online Rail Surface Defects Measureemnt Based On PMP

Rail is a quite important basic element in railway transport, supporting and conducting the operation of the train. In recent years, for meeting the request of the social development,railway is developing towards the direction of high-speed, heavy-load and high-density, which aggravates the abrasion and flaws on the rail surface. Running on the rail with surface defects,the train would not only affect passengers' comfort but bring great potential security threat.For the measurement methods on rail abrasion and surface flaws, it is still not reported about the 3D reconstructed of the rail shape and flaws, and the accuracy is generally not high for the additional ones. In this dissertation, online PMP (Phase Measurement Profilometry)technology is applied in rail detection, several key techniques are studied, the accuracy is pretty improved, and the 3D rail shape and surface flaws can be reconstructed. The main work of this dissertation is as follows:1. For in-service rail, the upper part may quite be smooth and shows high-reflective metal surface, thus the highlight of some fringe parts would cause CCD oversaturated, so as not to recognize complete deformed fringes. Considering this situation, based on the optical reflection character and phase change character of metal surface, the impact mechanism of the metal surface diffusion on the phase reconstruction is analyzed. The influence of wavelength,incidence angle, material and reflecting surface diffusion on the reflective light and phase change is simulated and analyzed, providing with the theoretical guidance.2. The mechanism of fringe distortion under divergent illumination with oblique projection is deduced and analyzed, and improved methods are given. The correction methods are simulated and errors are analyzed. Meanwhile, based on the homography matrix method,a fringe quality improvement method is presented to correct both aperiodicity and the heterogeneity of fringe intensity. In this method, the parameter of optical path is introduced to the mathematic model and the inverse projection is used for fringe generation. With this method, both the fringe period and intensity are homogenized to improve the fringe quality,and then the precision of online PMP is guaranteed.3. The four types of singular points in traditional Stoilov algorithm would cause nonsense points or super large error points in phase solution and unwrapping, thus further cause 3D reconstruction distortion or even reconstruction failure. In view of this problem, based on the fact that the probability of the true phase shift value is highest, the matrix of cosine square is processed,the singular points are ruled out one by one, and the most approximate phase value to the actual one is obtained with the probability and statistic. Experimental results show that this method could effectively rule out the singular phenomenon in traditional Stoilov algorithm,and thus improve the accuracy of the reconstruction.4. Aiming at another key technique- pixel match and combining with the rail online PMP measurement facts, the rail image parts outside the fringe are segmented for pixel match.Based on SIFT algorithm, feature points are sought and matched on this rail parts. And then,combining with image gradation detection, the remainder wrong matching points are further got rid of to realize the sub-pixel match.5. Several key techniques in vehicular online PMP are studied in this dissertation. Firstly,since actual reference plane are unavailable, a part of undeformed sine fringe is segmented from the gathered fringe image, on which the interpolation algorithm is adopted to extend to a virtual one as the reference plane, so that PMP can be achieved. Secondly, for the reasons of the vehicular detection platform shake, plus the mismatch of the vehicle velocity and the frame rate, etc. the gathered fringe images are often vague. Several algorithms are compared and analyzed in sharpening the vague fringe images. Thirdly, as rail is a long object while PMP's detection range is limited, the splicing and fusion of rail's 3D data are studied, hoping to show the measurement results of the whole rail line.6. The online PMP platform is built up in the laboratory and the system is calibrated.Based on the simulation experiments on the standard rail, factual flaws on the rail surface are measured and errors are analyzed. Results show that online PMP is a promising detection method on rail shape and surface flaws.