3D LiDAR Based Safe Driving Area Detection Method For Intelligent Vehicle

Environment perception is a key step for intelligent vehicles to realize unmanned driving.How to accurately represent the characteristics of three-dimensional scene has always been the focus and difficulty of the research in the field of unmanned driving.With the application of unmanned vehicle technology in increasingly scenes,3D LiDAR has been favored by researchers due to its high precision and rich information.Based on 3d lidar data,this paper researches the environment perception technology and safe driving area detection method of intelligent vehicles.The specific research contents include the following aspects:1.The background and research status of 3D environment perception are expounded.The principle and problems of 3D scene segmentation and intelligent vehicle travelable area extraction are analyzed.A safe driving area model is proposed,which reflects the safety coefficient of each driving region in a period of time and provides accurate data for intelligent vehicle path planning.2.Research on road detection methods based on lidar data.Aiming at the problem that the GPF algorithm is prone to mis-segmentation and poor real-time performance in the ground segmentation,an improved GPF algorithm is proposed.The accuracy and real-time of the algorithm is improved by using the grid height difference filtering and the block multi-plane fitting.On this basis,according to the difference of point cloud shape of the LiDAR raw data in different scenes,multiple features are constructed to extract road boundaries,and the accuracy of road boundary detection is improved by superposition of continuous data.3.Research on moving target detection and tracking methods.According to the target shape feature and the road boundary constraint,the moving target in the road are extracted,and a feature-assisted association method is proposed to correlate the moving target with the tracked target sequence.The Hungarian algorithm is used to optimize the correlation matrix,and the Kalman filter is used to predict the moving state of the target.4.An intelligent vehicle experimental platform was built,and a detection system for the safe driving area of the intelligent vehicle was designed and developed.A large number of experiments were carried out in the actual scene.The experimental results show that the system's function and performance achieve the expected goals,and also verified the effectiveness and real-time of the method in this paper.