Abstract: Aiming at the characteristics of point cloud of railway power line in airborneLiDAR (Light Detection and Ranging), especially when the horizontal projection was approximately overlapped, this paper proposed a method of automatic power line extraction.The paper used line fitting method, component analysis method and Hough transform method to calculate the direction of power line in the XOY plane, andused spatial distance clustering method to separate the two overlapped point clouds of horizontal projection, adopted the best clustering number judgment method of Davis-Bouldin indexto automatically calculate the number of power lines; applied the K-means clustering method to extract a single power line point cloud; and used the least square fitting method based on the quadratic polynomial to determine the parameters of power line model and carry out three-dimensional reconstruction.The experimental results show that this method can achieve high fitting accuracy.The fitting effect is good especially for the point cloud with approximate overlapping and interruption of horizontal projection.