Vehicle identification and GPS error detection from a LIDAR equipped probe vehicle

When detecting and tracking vehicles in LIDAR data it is important to differentiate between the measurements from vehicles and the measurements from fixed objects such as road boundaries. In this paper we present our approach to classify the LIDAR targets as vehicles and non-vehicle objects, and detect GPS positioning errors from a probe vehicle equipped with LIDAR and DGPS. The classification approach is developed based on a world space density image generated by averaging over historical LIDAR measurements. A correlation approach is used to detect GPS positioning errors and prevent incorrect classifications of the targets caused by erroneous projections between the vehicle coordinate system and the world coordinate system. The classification approach is then modified accordingly and becomes robust to these GPS errors