A Fusion of Dynamic Occupancy Grid Mapping and Multi-object Tracking Based on Lidar and Camera Sensors

Establishing a grid map containing dynamic and static information is an essential work for further research on motion planning systems that consider the interactive effects of multiple traffic participants. The algorithms will become efficient if we could take advantage of the interrelationship between dynamic occupancy grid mapping and multi-object tracking. Therefore, the purpose of this paper is to achieve a synergistic improvement in the effects of mapping and tracking algorithms by constructing the association between object tracking and map updating based on the fusion of Lidar and image information. After the fusion of the original Lidar point cloud and category information based on image deep learning, the static and dynamic grid regions in the grid map are updated separately. Among them, the particle filtering algorithm applied for dynamic grid optimization update utilizes the initial information given by the object tracking algorithm, and the update results of the dynamic grid in turn give guidance information for object tracking. This paper not only demonstrates the optimization effect of particle filtering on dynamic grid update when object tracking fails but also discusses the effect of the dynamic occupancy grid map on multi-object tracking accuracy and efficiency. The results show that the proposed method can achieve the establishment of the dynamic occupancy grid map and multi-object tracking simultaneously.