A Robust Vanishing Point Detection Method for UAV Autonomous Power Line Inspection

This paper presents a robust Vanishing Point (VP) detection method for Unmanned Aerial Vehicle (UAV) autonomous power line inspection. VP, an important visual cue for inspection navigation, is calculated by transmission lines. To achieve the robust extraction of transmission lines from the complex background, a neural-network-based line segmentation method (NNLS) is applied. From the segmented areas, further line extraction is carried out by Line Segment Detector (LSD) and Hough Transformation (HT). In addition, practical Random Sample Consensus (RANSAC) and line filter (LF) based on prior knowledge are adopted. Next the accurate vanishing point is calculated by linear least squares followed by Levenberg Marquardt (LM) optimization. For the verification of proposed method, an effective navigation model is developed. Finally, along with the constructed UAV platform, the entire system is tested on real inspection situations and achieves satisfactory results.