Automatic visual detection and verification of exterior aircraft elements

The French Air-Cobot project aims to improve preflight maintenance and provide a traceability of the performed checks. To do so, a collaborative mobile robot has been built. One of its non-destructive sensors is a pan-tilt-zoom camera. Visual inspection needs to be easy to apply and quickly carried due to time constraints on the airport. The robot is able to navigate autonomously around the aircraft, stop at checking positions and point the camera towards the region of interest. Items to inspect can be detected in the image acquisition thanks to some knowledge on those items. This knowledge has various forms: picture database, three dimensional models, aircraft markings, item configurations and marking dispositions. During its inspection tasks, the robot can warn its human operator that something is wrong and if possible ask for a fast human intervention for examples if there is still the chock in front of the landing gear or the protection on a pitot probe. At the end, it provides all its diagnoses and asks the human operator to validate or refute each of them. The article proposes fast visual detections and verifications of exterior aircraft elements. Previous researches focus more on static inspection, results in dynamic cases are also provided.