Shape And State Measurement Based On Machine Vision

Machine vision measurement technique has been widely used, it has many advantages such as high accuracy, non-contact, wide application and high degree of automation. In recent years, the industry of railway and subway is developed in a high speed in China. Railway vehicle wheelset plays key role in the safety of the railway transportation. So, it has active significance to monitor wheels to ensure safety and stationarity of the railway transportation. This paper based on the machine vision measuring technology takes railway wheel as the study objects to measure wheel size and detect the condition of the wheel tread automatically and on-line.At first, the dynamic wheel diameter measurement is studied in the paper. two kinds of measurement scheme for the diameter of the wheel are proposed, one is based on the extraction of two parallel structured light endpoints to calculate wheel diameter; the other uses the LED lighting from the side in order to reflect the wheel edge information then fitting the local edge profile to get the wheel diameter size. Compare the advantages and disadvantages of the two schemes. Then for0360 tread imaging and scratch detection, more camera acquisition system is constructed, study more image stitching, accurate tread image is obtained by image correction, edge fitting and neat, for image0360 stitching based on linear gradient fusion, the gray of tread image must be enhanced first.Through analyzing, image texture analysis method was adopted to detect defects of tread image, the experiment shows the variance, contrast and homogeneity are superior to describe the scratch areas,through K-means clustering scratch areas can be identified and defect levels can be defined. As strain is an important parameter in engineering application, strain measurement is explored by using the method of digital image correlation. The maximum strain can be got according to calculate the length of the virtual grid triangle before and after deformation, experimental results show the feasibility and accuracy of the method. Finally, the working process, software control interface and the client interface of the0360 tread imaging and scratch detection system are introduced. The strain measurement software system is developed by the technology of mixed programming of Matlab and C++. The two system has been used in the actual application, and they meet the requirements.