In view of the demand for high precision automatic extraction of power lines in the application of airborne laser radar technology, an improved algorithm for automatic extraction of power lines in complex terrain background is proposed. First, the ground and non-ground points are separated by the point cloud filtering algorithm for normalization, and the candidate point cloud is obtained based on the elevation characteristics of the power line point cloud and its continuous distribution in Z direction. Secondly, the Hough transform is improved by the peak enhancement local operator, and the precision extraction of the power line point cloud is realized. Finally, a single powerline can be obtained by the nearest growth algorithm based on the spatial continuity of the power line point cloud. The experimental results show that the extraction rate reaches more than 95% by using this algorithm. The algorithm can accurately extract power line points from the influence of terrain undulating and complex terrain, and can identify single line of power lines accurately.