image processing, navigation, obstacle avoidance, vanishing point, optical flow
Abstract
Indoor micro aerial vehicle (MAV) navigation is mostly based on high-computational cost obstacle detection algorithms. In this paper, we propose a path planning framework based on perspective cues. The vanishing points were detected by using the Hough-Canny transform. Due to the cluttering, multiple vanishing points candidates are arranged according to the proximity to the optical flow focus of expansion. Obstacles' depths are detected via the optical flow vector clustering. When multiple planes are recognized, the elected vanishing point is considered the center of the furthest empty plane from the camera. This remark guides the direction of the MAV to a particularly free-of-obstacles void area. Preliminary experimental results indicate that the proposed method is faster than other visual-based navigation algorithms.
