Power line detect system based on stereo vision and FPGA

Abstract

In recent years, Unmanned Aerial Vehicle (UAV) is widely used for power lines patrol. While it's more efficient than manually patrolling, the UAV may easily collide with power lines. Therefore, measures must be taken to detect the distance between UAV and power lines. This paper presents a real-time power line detect system based on binocular vision techniques, which contains a hardware platform based on FPGA to calculate real time disparity map. We adopt a robust stereo matching algorithm based on cross-based arbitrary shape support region method to calculate the depth image. As power lines in captured images are always too slender to be recognized in depth map, we propose a matching cost initial method based on morphology and mini-census transform to preserve the edge of the power lines in depth images. Our tests show that this new method can properly preserves the power lines of texture-less images in depth map.