A Multi-lidar Based Dynamic Obstacle Detection and Prediction Method

Moving object detection is a key issue in the dynamic environment perception of intelligent vehicles Intelligent vehicles drive in the outdoor dynamic environment,where the state is changing over time. The existing static environment map construction methods are limited in the dynamic environments. the detection is not enough for fast changing environment. And when an obstacle is a pedestrian or a vehicle, security becomes more critical. Therefore detection-and-prediction of trajectory move becomes crucial for robot cars.It has more important application value and greater application value for the dynamic environment to detect moving targets in the environment.This paper proposes a way to predict the movement of dynamic obstacles under automatic driving, using 3D lidar combined with laser sensor as the main sensor, and using it to detect obstacles in the direction of the vehicle. When obstacles are detected, DBSCAN is used to cluster the original data, which greatly improves the accuracy of clustering. Finally, the unscented Kalman filter is used to target to track.