Research On On-line Detection System Of Locomotive Wheel State Based On Machine Vision

For the safety of locomotive operation, it is very important to inspection the working situation of locomotive. However, the current artificial inspection brings high labor intensity and low working efficiency. Therefore, it is very critical for the locomotive wheel working to detect their working situation automation and intelligent. In the thesis, the wheel tread defect inspection and external diameter measurement were studied. By using image processing and optical measurement technology, machine vision based locomotive wheel tread defect inspection and external diameter measurement system was developed for the realizations of wheel tread damage automatic detection and dynamic external diameter measurement.The main work of the thesis includes as follows:1) For wheel tread damage detection, edge based detection and adaptive thresholding were firstly applied to preliminary determine the defect region. With texture characteristics of defect tread in the wheel image, the vectors of variance, contrast, inverse difference moment and correlation based on the gray level co-occurrence matrix were selected to describe texture characteristics of defecting tread in the wheel image. Combined with the gradient information of the cluster profile, K-Means clustering algorithm was used to locate and calculate the defecting area according to the similarity metric criterion. With the requirements of the system development, the defecting area is as the main index to judge final damage grade.2) For the wheel diameter measurement, the calculation method of diameter is determined by the relationship between circle and string. First, two parallel infrared laser line play in the outer side of the wheel, the wheel image with two strings was then obtained. According to the change of gradient in the region around the top-point or end-point of the laser line, local adaptive threshold is used to extract two laser lines on the wheel. At the same time, according to the connectivity of the laser line, the top and end points of the laser line were finally detected. At last, these four points on the two strings extracted were used to calculate the external diameter of the wheel.3) A defect detection and external diameter measurement system baesd on machine vision for locomotive wheel tread was developed. With the requirement of the performance of system developing, the hardware and software of the system were designed in which the locomotive wheel dynamic acquisition, wheel tread automatic extraction, tread defect online inspection and defect grade intelligent judgment, wheel tread image splitting and online data real-time transmission were achieved. The developed system is carried out in on-track detection testing field in North America TTCI. The experimental results show that the system can complete online wheel defect inspection and defect grade identification. The detection accuracy of defect over two grade reached 90%.