Crack Inspection of Wooden Poles based on Unmanned Aerial Vehicles

Abstract

Wooden poles are deployed for decades. To inspect the poles, visual inspection by human experts is generally adopted. However, poles are mostly deployed in areas with bad traffic conditions, it is hard to access them. Furthermore, human inspectors tend to get tired, which lowers the inspection accuracy. In this paper, we present an unmanned aerial vehicle (UAV) based system for automatic wooden pole inspection. First, we captured the pole images by using a drone with a high resolution camera, totally we captured 600 pole images. Second, we proposed a encoder-decoder based neural network that can segment poles and cracks with high accuracy. Finally, we designed a processing scheme that can give the location and direction of each crack with high accuracy.