Power Line Extraction Based on Combined Clustering of Line Segments

Line segment detection based on Hough transform is widely used in Power Line Extraction (PLE). Because power lines have the characteristics of symmetrical edges, there are many problems of multiple detection and poor positioning in PLE using Hough transform and improved line segments methods. In order to solve the above problems, A novel PLE method based on combined clustering of line segments is proposed. First, Image graying, denoising, enhancement and edge detection methods are selected to weaken the effects of complex background around power lines; second, convert the straight line in Cartesian coordinates into points in polar coordinates, determine the power line segments by accumulator; third, determine the number of power lines by DBSCAN algorithm which is insensitive to abnormal points, then cluster the power lines by K-means algorithm, and finally convert the power lines from polar coordinate space to Cartesian coordinate system. Experiments show that the proposed method can automatically extract power lines and effectively solve the problems of multiple detection and poor positioning.