Sensor fusion with Occupancy Fusion Map for pedestrian detection in outdoor environment

An occupancy fusion map (OFM) approach with sensor fusion for pedestrian detection is described in this paper. This work emphasizes the idea of redundancy or/and complementary due to the different data information provided by the laser scanner and stereovision. Two approaches are presented to achieve the robust pedestrian detection result: (1) an novel method of building occupancy grid map with stereovision to solve the problem that the independent object cannot be detected because of discontinuous occupied data in traditional X-Z gird map; (2) the approach of tracking to solve the continuity assignment of target objects, can also resolve the problem of two peoples are linked together and cannot be detected respectively. Experimental results showed promising performance of our approach.