Object Detection And Tracking For Intelligent Vehicle Based On Multi-sensor Information Fusion

Intelligent vehicle refers to a new type of vehicle that integrates various technologies and equipped with advanced on-board sensors,controllers,actuators and other devices,which can realize information exchange to achieve safe,efficient and energy-saving driving and ultimately realize automatic driving.Environmental perception technology mainly refers to the detection of the surrounding environment of the vehicle based on on-board sensors,which is an important guarantee for the driving safety of intelligent vehicles.In order to ensure the safe driving of intelligent vehicles,the stability and reliability of environmental perception system need to be guaranteed.Therefore,it is necessary to detect and track objects in the driving area of the intelligent vehicle.Based on the intelligent driving multi-disciplinary platform of Tianjin University,this paper conducts a scheme layout of the sensor platform for different application scenarios,and realizes object detection by using on-board lidar.This paper also propose a multi-sensor information fusion object detection and tracking scheme for the intelligent vehicle environmental sensing system and using lidar and camera to detect and track objects in passable area.The main contents are as follows:1.The sensor layout scheme of the intelligent driving multi-disciplinary platform of Tianjin University was designed.The sensor model was investigated according to the scene needs and the requirements of information fusion.The location layout was conducted according to the sensor model.2.The on-board lidar was used for object detection and the lidar data collection,passable area extraction and point cloud clustering were completed according to the system requirements,and the DBSCAN algorithm of point cloud clustering was optimized.Finally,raster map was generated according to the point cloud clustering results.3.An object detection and tracking method of information fusion is proposed according to the data types of lidar and camera.The scanning point cloud data is projected onto the picture through the joint calibration of lidar and camera to achieve information fusion,and particle filtering algorithm is optimized to achieve object tracking.The validity of the algorithm is verified by using KITTI data set.