| Wheel of train is important moving part, and it plays an important role on the train traffic safety. Detection for the size of Wheel of locomotives in China is still in common use tools such as caliper measurement, this traditional manual measurement of wheel on detection method has low precision, high error rate of measuring and low efficiency. Therefore, in order to quickly accurate measurement of train wheel size, to ensure that the wheel wear amount does not exceed the prescribed limits, based on this basis, this paper puts forward the application method for the measurement of wheel dimensions and light shots using the method of curve fitting wheel, having very important realistic significance for ensuring the safe operation of trains.This paper mainly studies a kind of wheel wear based on machine vision automatic detection system. This system mainly is the laser as a light source to the wheel tread of scanning, thus forming a band of wheel, at the same time with high precision linear CCD sensor for real-time acquisition of the band profile shooting. Then the image acquisition and image analysis by computer, contour extraction and refinement of wheel, and standard contour line matching, accurate and fast detection by flange and tread wear. This paper focuses on the calibration and image detection and elaborates the system, and the system calibration and image detection of various algorithms are analyzed. System calibration is used for Tsai two steps calibration method, improving its further to improve the accuracy of calibration. Image detection by the original image of wheel profile are image preprocessing, image feature information extraction, sub-pixel edge detection and thinning contour analysis process, improving the quality of the image, to facilitate the subsequent analysis and calculation. The principle of the filter are analyzed and summarized, the reasonable selection of Gauss filter to suppress noise in the image; the principle of the gray scale changes, image histogram equalization and threshold segmentation are analyzed, and the wheel image preprocessing using the algorithm; principle of many operators of edge detection and sub-pixel detection of analysis, the reasonable choice of sub pixel edge location of wavelet transform and Canny edge detection operator and the calculation of the first order moment of the pixel level algorithm extracted refinement on the wheel on image, image information extraction accuracy is further improved.This paper has improved the detection accuracy of wheel, through the analysis of field measurement and the system measured data, error of wheel set wear volume within0.2mm, meeting the actual requirements. There are58charts,4tables,54references. |
