A robust dynamic programming algorithm to extract skyline in images for navigation

In this paper, a skyline image detection algorithm is proposed for navigation of mobile vehicles or planes in mountainous environments. First, edge detection and subsequent binary thresholding are applied on the luminance component to obtain the edge map, from which a multi-stage graph is constructed for determining the skyline curve by using the dynamic programming (DP) algorithm. In optimal DP search, characteristics (e.g., preferred position and orientation) of the skyline are utilized to help in correct linking of curves. The tolerance in short breakage of skyline curve is considered for robustness consideration. Experiments show that the processing speed is fast (approximately 0.12-0.21 s for a 352 × 240 pixel image on a Pentium-M 1.3 GHz CPU) and promising for real-time applications.