【摘要】 为解决钢轨轮廓检测中标定平面和激光平面不重合、钢轨轮廓难以动态匹配的问题，制作了一种阶梯型标定板，并提出基于Harris-SIFT和改进RANSAC算法的钢轨轮廓匹配方法。首先设计制作一种阶梯型钢轨轮廓标定板，能直接标定激光平面，避免了实际应用中标定平面与线激光平面不完全重合的问题。标定过程中采集最佳控制点，通过多项式变换和双线性插值进行图像校正，然后通过Harris-SIFT算法进行轮廓粗匹配，大大缩短了匹配时间，最后利用改进的随机抽样一致算法(RANSAC)进行轮廓误匹配滤除，实现了轮廓精匹配。实验结果表明，图像标定时，相对于棋盘格标定板，校正误差减小0.042 mm;轮廓匹配时，相对于传统轮廓匹配方法，改进后的轮廓匹配算法匹配运行时间快，正确率达100%。更多还原

【Abstract】 In order to solve the problem that the calibration plane and laser plane do not coincide, and the rail contour is difficult to be matched dynamically in the rail contour detection, stepped calibration plate is fabricated, and a rail contour matching method based on Harris-SIFT and improved RANSAC algorithm is proposed.Firstly, a step type rail contour calibration plate is designed and manufactured, which can directly calibrate the laser plane and avoid the problem that the calibration plane and the line laser plane do not coincide completely in practical application.In the calibration process, the optimal control points are collected, and the image is corrected by polynomial transformation and bilinear interpolation.Then, the Harris-SIFT algorithm is used for rough contour matching, which greatly reduces the matching time.Finally, the improved random sampling consistent algorithm(RANSAC)is used to filter the contour mismatching, and the contour fine matching is realized.The experimental results show that the calibration error is reduced by 0.042 mm when the image is calibrated compared with the checkerboard calibration plate.Compared with the traditional contour matching method, the improved contour matching algorithm has a faster matching time and 100 % correct rate.更多还原