Automatic Detection Of Wearing Train Wheel Tread Based On CCD

Railway transportation has already become one of the important traffic tools in the field of cargo transportation and travel in our country. With the development of technology and manufacture quality, the speed of train has become more and more faster. Therefore, in repair factory not only the requirement of manufacturing quality and accuracy become more and more stricter, but also the requirement of detection precision and repair precision become more and more stricter. Because of the prolonged contact friction with pathway and slowly corrode with water and oxygen in the outside world, the train wheel will inevitably become wear and damage. The phenomenon of wear and damage will not only influence the service life and replacing frequency of a train wheel, but also increase the potential risk in the period of train running.The traditional artificial measurement is still the most commonly used measurement of detection in the most of repair factory of train wheel in our country at present. The 2D measurement based on cross direction section cannot achieve the requirement of accuracy on the wear of tread circumference of a circle and the thickness and height of rim. And the information which be measured by traditional artificial measurement is incomplete, cannot express the information of the whole wheel. In order to make up various insufficient of traditional artificial measurement, image processing technology which combined with high performance computer is used to detect tread wear of train wheel automatically and determined the economics plan of lathing of train wheel in this paper. It provides new ideas and methods in the field of automatic detection and repair of train wheel.According to the requirement accuracy of binocular vision automatic detection of wearing train wheel treads, a series of hardware equipment were chosen, such as lighting source, optical lens, array CCD camera and image acquisition card. And the binocular image acquisition systemwhich was suitable for wheel was built. Firstly,Zhang Zhengyou’s calibration was chosen to obtain the internal and external parameters of binocular camera. And image gray processing, image smoothing, image enhancement and image sharpening were successively used to improve the train wheel binocular image.Then, SIFT algorithm and constrain criterion were analyzed in the process of feature matching. Aiming at the image feature of wearing train wheel tread, an improved SIFT algorithm using cosine distance to match feature in a limited range was presented and feature of wheel tread was extracted and matched rapidly.Finally, ICP algorithm was used to fuse the point cloud of train wheel tread, the equation of train wheel axis was obtained by point cloud and the constraint relation in train wheel. The 3D model of wearing train wheel tread was reconstructed and the wear model was compared with standard model in the same axis, then the wear condition of train wheel tread was obtainedand the economics plan of lathing of train wheel was determined.