Object Detection and Tracking through Non-Linear Filtering and Fusion of Lidar and Position Data

This article presents a new method of object detection and tracking in the context of driver assistance systems and autonomous automobiles. Two lidar scanners that scan the environment periodically serve as the data source. The sensors' raw data is then fused with high precision positioning data in order to transform it into a global coordinate system. Then, the transformed data is used to feed the particle filter. Within this filter, an importance function evaluates the newly gained lidar points to weigh the individual particles. Finally, a cluster analysis in the observation space extracts the estimated object states.