基于 LoG-Zernike 矩算子的钢轨激光图像提取识别

摘要

In laser and photogrammetry techniques for track profile measurement,the extraction accuracy and efficiency for the laser stripe centerline are two key issues.It is proposed to utilize LoG-Zernike moment operator to guarantee the localization precision.LoG operators are applied to the first and rapid localization of the track profile edge,then in the neighborhood area of the image edge,Zernike moment operators are used for the further processing,achieving the subpixel extraction precision for the track laser profile images.In this paper,the principle of LoG-Zernike moment operators for subpixel extraction is explained in detail,and LoG-Zernike moment operator computational formulas and steps are carried out.The extraction lines for the laser profile image of tracks,processed by LoG and LoG-Zernike moments operators respectively,are compared.The results show that LoG-Zernike moment operator extraction line is thinner and more continuous. Additionally,LoG-Zernike moment operators are used in the current track inspection vehicle for image processing,and the experiment demonstrates this operator is practicable.%采用激光摄像技术进行钢轨轮廓测量，图像中钢轨激光光条中心提取精度及效率，是决定钢轨轮廓测量是否可行的关键因素。提出采用 LoG-Zernike 矩算子进行钢轨激光光条中心提取，先利用 LoG 算子进行钢轨激光光条边缘粗定位，在粗定位图像边缘邻域内，再采用 Zernike 矩算子进行激光光条中心精确定位。详细论述了 LoG-Zernike 矩算子进行图像亚像素提取原理、过程，给出了 LoG-Zernike 矩算子详细计算步骤。试验结果表明，LoG-Zernike 矩算子提取的钢轨轮廓曲线较细，连续性较好。