Object Detection And Tracking In Driveable Area Based On Four-Line Laser Radar

Intelligent vehicle,an important part of intelligent transportation system,is mainly to solve the increasingly serious traffic problem.It includes environmental perception,planning decision,control execution and so on.Thus the performance of environmental perception affects directly the safe driving of the intelligent vehicle.In this dissertation,object detection and tracking in driveable area are studied based on the active sensor four-line laser radar.The main contents of this dissertation are as follows:(1)An improved density based spatial clustering of application with noise clustering algorithm is proposed,which based on the working principle and data analysis of the laser radar.According to the characteristic of laser radar scanning to structured or semi-structured road,the k nearest neighbor method is used to draw Dist_k.The algorithm parameters(Eps,Minpts)can be adaptively selected according to the characteristics of the hierarchical data set and complete the muti-density clustering.The improved algorithm can accurately cluster the laser radar data points with different density levels,which lays the foundation for the road information extraction and obstacle detection.(2)Road information extraction method based on the characteristics of the laser radar scanning to the road is proposed.The second extraction algorithm based on collinear points is proposed which can accurately extract the curb data.We obtain the height of the curbs,eliminate interference curb.Finally,curbs fitted out by least square method.Then the road surface data is marked,and the information of obstacles is extracted from the remaining data,such as distance,position and dimension etc.The experimental results show that the proposed algorithm can be used to extract road information in complex road scenes.(3)The grid map is built directly in Cartesian coordinate system by using the inverse sensor model,and the updating algorithm based on Dempster-Shafer theory(DST)and Proportional Conflict Redistribution rule 2(PCR2)is proposed.The aim is to solve the poor fusion result of DST in high conflict.The coordinate transformation is not needed which improves the translation accuracy and the calculation amount decreased.The conflict information is used to detect the moving objects.Then the closed operation and the priority marking algorithm based on regional growing are used for the cluster analysis,the information of the moving object is extracted.(4)The detected moving target is tracked in the occupancy grid map.Data association algorithm is studied of muti-target tracking,and an improved joint probabilistic data association algorithm combined with Kalman filter is proposed.In addition,a variable tracking gate for each tracked target is built,which increases the adaptability of the moving target tracking system to the environment.The proposed algorithm can exactly matching the optimal target in dense environment and track moving objects stably and accurately.