基于单目视觉的轨道固定桩基准点测量方法
Measurement method of reference points of railway fixed pile based on monocular vision

为了提高铁路线路固定桩基准点绝对坐标测量的效率,提出了一种新的测量方法; 建立了无合作目标的单点测量模型,采用单目相机采集激光靶标图像,利用光饱和点重心法提取激光光斑中心; 研究了平面直线成像规律,构造了基于正交直线的单应性矩阵求解方法,并对图像进行透视畸变校正; 根据校正后的图像与靶标的几何相似关系,计算了激光光斑与靶标的横、纵向偏差; 在室内环境下,进行了靶标图片拍摄的正交试验,计算与比较了横、纵向偏差。试验结果表明:在激光光斑和靶标固定的条件下,保持相机与靶标的距离不变,改变相机角度拍摄图片,经过透视变换校正后,横、纵向偏差与期望偏差分别为0.082、0.254 mm; 相机拍摄角度固定,改变相机与靶标距离拍摄图片,经过透视变换校正后,横、纵向偏差与期望偏差分别为0.126、0.014 mm; 在相机的角度、相机与靶标的距离都改变的情况下,拍摄的图片经过透视变换校正后,横、纵向偏差与期望偏差分别为0.329、0.064 mm; 可见3组试验的横、纵向偏差与期望偏差的误差均小于0.5 mm; 系统的水平距离测量误差范围为±1.52 mm,高程测量误差范围为±0.67 mm,根据轨道检查仪性能指标,线路水平距离误差范围为±3.0 mm,高程误差范围为±2.5 mm,因此,本文的测量方法精度满足轨道测量要求。水平距离测量误差完全由激光测距仪和倾角传感器决定,而高程测量误差是由激光测距仪、倾角传感器与激光点和靶心的偏移量共同决定的。
To improve the efficiency of absolute coordinates measurement of railway line fixed pile reference points, a new measurement method was proposed. The single-point measurement model of non-cooperative target was established. The laser target images were captured by monocular camera, the centroid method of saturation point was applied to exact laser spot center. The imaging rules of lines in plane were studied. A homography matrix solving method was constructed based on the orthogonal straight lines, and the image was corrected by perspective distortion. Then the horizontal and vertical offsets of laser spot and target were calculated based on the geometric similarity relationship between corrected image and target. In the indoor environment, orthogonal experiments of target image shooting were carried out. The horizontal and vertical offsets were calculated and compared. Experimental result shows that under the condition that laser spot and target are fixed, the distance from camera to target is constant, and the camera's angle is only variable, after the images are shot and corrected by perspective transformation, the errors between the horizontal and vertical offsets and the expected values are 0.082 and 0.254 mm, respectively. Under keeping the camera's angle and changing the distance from camera to target, the errors of corrected images are 0.126 and 0.014 mm, respectively. When the camera's angle and the distance from camera to target both change, the errors of corrected images are 0.329 and 0.064 mm, respectively. The errors between the horizontal and vertical offsets and the expected values of 3 tests are less than 0.5 mm. The range of horizontal and height measure errors of the system are ±1.52 and ±0.67 mm, respectively. According to the index of railway inspection tester's performance, the horizontal measure errors of railway are within ±3.0 mm and the height measure errors are within ±2.5 mm. Obviously, the precision of