A track-before-detect approach for extended target tracking in multi-lidar systems using a low-level centralized fusion framework

This paper presents a new method of multiple object tracking in a multisensor lidar context. One of the difficulties of scanning laser range finders (LRF) lies in the measurements of the objects since they are spatially distributed. We propose a tracking method (MF-IP-SMC), which uses the LRF data from a clustering stage (IP method) which avoids the usual detection step. This step is indeed source of detection problems or false alarm. The MF-IP-SMC approach is then validated on multilidar sequences of vehicles on road. A comparison with reference methods in vehicle tracking is then presented in order to quantify the contribution of the proposed solution.