[Talks@IDSIA] Joachim Ott, ETH: Vision­-based surface classification for micro aerial vehicles

Short talk tomorrow (Wednesday May 27) in the meeting room at 11:30 by Joachim Ott, a student at ETH working with the Robotics and Perception Group at the University of Zurich. Title: Vision-­based surface classification for micro aerial vehicles Abstract: Micro Aerial Vehicles (MAVs) are becoming an increasingly important tool assisting humans in tasks like rescue and monitoring. To succeed they have to be able to navigate autonomously using vision and inertial measurement as sole sensor modalities, i.e. no GPS or laser scanner. Our current Robotics and Perception Group framework allows 3D reconstruction of the MAVs environment by using just a single camera. This reconstruction enables an autonomous landing procedure, during which the MAV scans the ground for a flat surface and subsequently initializes landing. The current project aims at extending the landing procedure with a surface classifier, where camera images as well as an elevation map are used to classify the ground into classes (e.g. grass, gravel, water etc.) and assigning them a landing score, i.e. how preferable it is to land on it.