POWER LINE DETECTION BASED ON MAXTREE AND GRAPH SIGNAL PROCESSING

Low-altitude unmanned aerial vehicle (UAV) remote sensing facilitates the frequent detection of power lines and liberates manual inspection. In the process of UAV power line inspection, power line detection in UAV aerial images plays an important role. But false alarms and miss alarms often occur during the detection process. Aiming at this problem, a power line detection method based on Maxtree is proposed. This method transforms the UAV aerial images into a graph structure, i.e., Maxtree, and detects the power lines under graph signal processing frame. Two-stage filtering is designed to preserve power line components. The preprocessing stage filters out most of the background part according to the color value, and the other stage performs filtering on the Maxtree created with connectivity and gray value. For each node, three attribute components, i.e., gray value, linearity, and length, are assigned to facilitate power line detection. Experiments show that the method can detect power lines accurately and effectively.