First-look enabled Autonomous Aerial Visual Inspection of Geometrically Fractured Objects in Constrained Environment
Vignesh Kottayam Viswanathan, Gajanan Satpute. Sumeet, Björn A. Lindqvist, George Nikolakopoulos · Zenodo (CERN European Organization for Nuclear Research) · 2022
In this article we propose a novel offline, model-based aerial visual inspection scheme for geometrically fractured large objects, based on fully autonomous Unmanned Aerial Ve-hicles (UAV s), while specifically targeting the case of constrained environments. The proposed framework enables a safe and collision free inspection mission, while guaranteeing a complete visual inspection of the object of interest. The proposed framework employs a novel First - Look approach to generate viewpoints satisfying specific photogrammetric requirements, as well as spatial constraints that are inherently applied by the UAV's state constraints. As it will be presented, i) the First - Look approach allows the UAV to first orient it's view vector towards the nearest available point detected by kd - tree based Nearest Neighbour search on the object, from it's current position, and ii) in the sequel, based on the orientation of the left vector of the camera and the overlap distance, the next viewpoint is projected. This approach is repeated throughout the whole inspection procedure, while the established framework has also the merit to ensure that the inspection path adapts to the shape of the object, which is highly advantageous for the cases of geometrically fractured objects. Multiple realistic and physics based simulation results are presented that prove the efficacy of the proposed scheme.