Semantic classification of urban 3D point clouds

An autonomous vehicle (AV) requires an accurate perception of its surround- ing environment to operate reliably. This perception system transforms data from multiple sensors into semantic information that enables autonomous driving. Road detection is essential for AVs to locate themselves and make...

Πλήρης περιγραφή

Λεπτομέρειες βιβλιογραφικής εγγραφής
Κύριος συγγραφέας: Παπανδρέου, Ανδρέας
Άλλοι συγγραφείς: Papandreou, Andreas
Γλώσσα:English
Έκδοση: 2020
Θέματα:
Διαθέσιμο Online:http://hdl.handle.net/10889/14370
Περιγραφή
Περίληψη:An autonomous vehicle (AV) requires an accurate perception of its surround- ing environment to operate reliably. This perception system transforms data from multiple sensors into semantic information that enables autonomous driving. Road detection is essential for AVs to locate themselves and make a rational decision, especially under obstacle occlusions, road discontinuities, and curved road scenarios. We are attempting to solve the above challenges implementing geometric approach algorithms rather than machine or deep learning techniques. A robust method for road curb detection is proposed, by correlating the outputs of different perception modules, through a temporal correlation between consecutive frames. These modules consist of the sensors of Light Detection and Ranging (Lidar), Global Positioning System (GPS) and Inertial Measurement Unit (IMU). As soon as the road has been detected, a 3D point cloud semantic classification approach using multiscale features with a consistent geometrical meaning is proposed to detect the objects that exist above the road. Finally, several experiments have been conducted using data from both real and simulated environments.