Detection Techniques Based On Image Processing And Machine Vision Vehicle Wheel Tread

Wheelset is an important running part of the railway vehicle, and the damage of the wheelset tread will cause many problems and impact the safety of train transportation badly. Therefore, the damage should be detected quickly and accurately to ensure its wears under the limitation for the stabilization and safety of railway transportation.Nowadays the tread of vehicle wheelset is still measured mainly mechanical tools in our country, the measurement method is low efficient, low reliable and inconvenient for information management.The technology based on image treatment and machine vision for wear detection and correlation parameters measurement of the wheelset tread is presented in this paper. While a wheelset is rolling under a certain speed controlled by step motor, tread surface of the wheelset is scaned by the laser, and the light band profile is collected by array CCD(Charge Coupled Device) camera. The collected image can be processed and analyzed by image type transformation, image anti-color, median filtering, LOG(Laplacian of Gaussian) sharpening operator, threshold segmentation which based on iteration, filament arithmetic etc, and these methods make the extracted image information reach relatively high accuracy. The calibration of system is realized by two-step method, which also corrects distortion. The wear of wheelset tread is detected quickly and accurately by different shadow arithmetic of pattern recognition, and correlation parameters of the wheelset are measured by image addition operation. Finally, error sources of experimental results are analyzed and a specific program to improve the measurement device is put forward.