Robust and Efficient Pose Estimation of Pipes for Contact Inspection using Aerial Robots

This work describes the methodology for detecting pipes and their pose in refineries inspection using Unmanned Aerial Vehicles (UAV s) for remote Ultrasonic Testing (UT). Segmentation techniques such as the Hough Transform and its variations, and Random Sample Consensus have been widely used. This paper is therefore focused on the development of an efficient computer vision algorithm to detect the position and orientation of the pipes in order to land on them autonomously to perform the inspection, by using 3D point cloud information from depth cameras. Applying a methodology based on Random Sample Consensus and point cloud pre-processing to fasten the algorithm performance has led to robust estimations of the pipes and their poses in an indoor testbed using a realistic environment, allowing the autonomous landing and the subsequent inspection.