Research on Non-uniform Grid Map Building and Object Detection Technology Based on LiDAR

Occupancy grid is a practical method for environment map building. A drawback of occupancy grid is the high computational cost, the smaller cell size the higher map accuracy, but the overall number of cells increase quadratically, meaning that the rising computational cost, thus a trade-off between the grid map accuracy and the computational cost is required. To address this issue the non-uniform occupancy grid is used for environment modeling and object detection in this paper, which not only reduces the number of cells but also achieves high accuracy grid in the vicinity of ego vehicle. Then, according to the distribution characteristic of the non-uniform grid, the obstacle is extracted with the distance-based cluster algorithm. Finally to fit the obstacle with a 3D bounding box. Experiment show that the memory usage of improved method reduces by about 81% and 86% respectively, the time consumption for building the grid map reduces by about 72.37% and 73.72% respectively.