Abstract: Aiming at the problems of difficulty in surveying and mapping of existing railway line, low operation efficiency and insufficient measurement accuracy, this paper proposes a high-precision reconstruction algorithm for existing track lines based on the principle of unmanned aerial vehicle (UAV) multi-view geometry. Firstly, extract the straight line segment from the UAV images. Then, the image-based measurements are used to obtain candidate matching lines, and Powell optimization based on the distance are performed to obtain accurate rail line matching results. Finally, using the image rail line matching results as the observation value and the reprojection distance of the rail lines is used as a measure, the coordinate of rail straight line segments is optimized by the least square method. Experimental results show that the proposed method realizes automatic image rail line matching and high-precision adjustment solution. The image-based reprojection residual is better than 0.5 pixels, and the internal coincidence accuracy of railway gauge and rail height are 0.006 m and 0.005 m, respectively. The horizontal and vertical accuracy of the proposed method can reach about 2 cm, and the adjustment model has good convergence, which could meet the requirements of engineering application.

Key words: UAV, multi-view geometry, line feature, least square optimization, existing railway