The Research On Locomotive Wheelset Tread Scrape Image Inspection Algorithm

Wheelset is one of most important parts of locomotive running gear. Its working condition directly affects locomotive’s safety. Tread scrape is a key factor that influences working safety, stability and economy of running. Thus, inspection of tread scrape becomes an issue that payed more attention to and improved constantly by stuffs of railway department.As one of the important ways of non-contact measurement, the image processing can effectively improve the quality of the original image and obtain the main feature of target. It is well known that a good image processing method can get good measuring precision. Non-caontact with measured object, image processing has advantages as follows:safe and reliability, high measurement speed, facility of realizing automatic measurement, little effection by external environment etc. Therefore, the automatical, fast and precise inspection problem of locomotive wheelset tread can be resolved.Firstly, image capturing scheme is proposed in the paper by researching the domestic and international popular image capturing systems. And locomotive tread scrape images are shooted in the condition of both natural light and active lighting of darkroom with the resulting of better definition of the images and perfect charactersistics of feature region.Next, actual captured wheel scrape images are processed and analylized with image processing algorithms. By using these algorithms, wheel scrape in the images are positioned and measured precisely. The main algorithnms include image graying, filtering, boundary enhancing, feature points projection, regional segmentation, target positioning and so on.In addition, in the case that actual wheelset images is not enough, the method of3-D computer simulation is introduced to provide various tread scrapes for iamge processing experiment. Fidelity of simulated images is ensured by entity render settings with CAD. Property of image processing algorithnms is further validated by processing these images.Finally, MFC interface is adopted, and part of the processing algorithnms is embeged into the interface. Windows in the interface can be updated in realtime, which provides convenience for processing results observation and comparison.