The On-line Image Detection System Of Wheel Tread Of Train

In recent years, with the rapid development of high-speed rail and rail transit in China, how toprotect their safety became a major issue of social concern. The train wheel tread abrasion operationalsafety is an important factor. Currently most of the scratches on the wheel tread damage detections arestill static detecting, which can‘t meet the actual needs. But compared to static detection systems,on-line testing is a high degree of automation, saving human resources and time.In this paper, through research the existing abrasions detection systems at home and abroad, awheel tread scratches (dynamic) on-line detection system based on linear CCD cameras wheels isproposed. The system utilizes relative motion between a stationary linear CCD cameras and themoving locomotive to image the wheel tread and get information of tread damage by imageprocessing. Besides a CCD camera, the system also includes a light source, industrial controlcomputers, speed and trigger sensors and software system. In the working process of the system,firstly, the train goes through the trigger sensors, then the camera are started to get ready to captureimages start, while the speed sensor will turn the train speed to frequency as the camera acquisitionfrequency.When the train wheel reaches the appropriate position, the trigger fires the camera to startscanning wheel tread, then the images are saved in the industrial control computers.This paper does numerical design on the work processes and architecture of the systemtheoretically. The system requires multiple cameras to complete the360-degree wheel tread imageacquisition. First the parameters in the system are calculated to determine the adjustment range of thevarious imaging parameters. On this basis, through the comparative analysis of the experimentaleffects of various parameters on the system imaging, the best system layout parameters can bedetermined. Then image correction, matching, enhancement and recognition processing are used inthe collected images to get tread damage images. Since there‘s deformation in the images collected bycameras, so the image correction is used; and the information of a wheel tread is composed of imagesscanned by three cameras, the images are still need splicing. Finally, multi-threading and MATLABand C++mixed programming are used to develop the system software. Installation andcommissioning of the actual experimental results show that the system can work effectively.