This document is an expanded abstract.
This paper presents a control mechanism based on the actual distance between the Unmanned Aerial Vehicles (UAV) and the tracking structure for UAV visual tracking of pipeline on the ground in a low altitude. The technique includes two parts: structure identification and navigation control. The unburied pipeline is considered as a simple line in 2D image, then the Canny Edge Detector (CED) and Probabilistic Hough Transformation (PHT) are used in image processing to detect edges, identify the pipeline and extract useful parameters (angle and distance in image). For autonomous tracking, the response time, flight stability and tracing accuracy are influenced by the navigation and control performance severely. Thus in this paper the position controller is designed which can calculate the actual distance between the UAV body frame and the detected structure, and the UAV position is regulated by the variant PID controller until the UAV is flying exactly over the pipeline on ground.
In the last few decades, UAVs are starting to be used in civilian operations especially in industries. Aiming at improving the efficiency of work and reducing cost of companies, increasing number of UAVs are used to accomplish these dangerous and difficult jobs. So far autonomous visual tracking of linear structures by UAVs has an increasing attractiveness to the public for its extensive industry applications, such as traffic tracking (B. Coifman, 2004), electrical cables inspection (D. Jones, 2005) and pipeline inspection (A. Shukla, 2016). Specifically, for ensuring the normal supply of electricity, gas and oil, these facilities are generally inspected from the air due to their extensiveness. As known vision is one of the most important part of human beings, so does it for UAVs. For example, UAVs can reach obstacle avoidance by recognizing obstacles position and calculating the distance via robot vision. Without Robot Vision some robots can also work well, but for a number of specific applications the blind robot is helpless.