A perception system for obstacle detection and tracking in rural, unstructured environment
Yechuan Yeo, Xinghu Zhang, Rong Yang · International Conference on Information Fusion · 2014
This paper describes a multi-sensor perception system for detecting and tracking obstacles in a rural, unstructured environment. A 3D lidar is used as the main sensor for obstacle detection and short range vehicle tracking, complemented by automotive radars for long range moving vehicle tracking. A bottom-up inference process based on extended elevation map is used for obstacle detection, while a modified random matrix method is adopted for laser-based vehicle tracking to handle rigid body constraint and partial observability problem. Track-to-track level fusion is performed to improve the accuracy of the individual sensor's tracks. The performance of the system is evaluated and verified through extensive experiments conducted out-field.