Research On Detection Method Of Driving Area And Obstacle Based On Four-layer Lidar

With the rapid development of economy and communication level,driverless car,as an important direction of future traffic development,has gradually become a hot topic at home and abroad.The research topic of driverless car covers many aspects of knowledge system,the most important four aspects are: positioning,perception,decision-making and control.Four-layer LIDAR is widely used to realize the relevant functions of environment perception of driverless car because of its advantages such as moderate amount of data,high accuracy of measurement,multi-target detection and little influence of environmental factors.In this paper,the main research contents of driverless car using four-layer LIDAR are as follows:(1)When the driverless car detects the driving area,the clustering effect of density based clustering algorithm is poor because the data density of LIDAR varies unevenly.An improved Jarvis Patrick(JP)clustering algorithm is proposed.The algorithm measures the local density of data by the relationship between k-nearest neighbor and shared nearest neighbor,and selects representative points,which has scalability to the change of data density,thus increasing the search speed and accuracy of the algorithm in detecting the driving area.The improved algorithm can improve the accuracy of clustering the data with different density.(2)On the basis that the improved Jarvis Patrick(JP)clustering algorithm can adapt to the uneven change of LIDAR data density,this paper proposes a detection algorithm of driving area,which extracts road edge clusters from the improved JP clustering clusters by analyzing the data of different objects in the vehicle environment scanned by LIDAR.In order to reduce the influence of some noise data in the road edge cluster and obstacle data outside the road edge on the road edge fitting,random sample consensus(Random Sample Consensus,RANSAC)algorithm is used to fit the data points in the road edge cluster to get the road edge in front of the driverless car.The drivable area is detected by combining the width,position of the obstacles in the road ahead with the width safety factor of the driverless car.(3)In order to solve the problem of repeated neighborhood search in the traditional connected area marking algorithm when detecting the dynamic obstacles in the driving area,a connected area marking algorithm based on the grid map established by D-S evidence theory is proposed.The real-time performance of the dynamic obstacle detection algorithm is enhanced by reducing the number of repeated neighborhood search.(4)In view of the false detection phenomenon caused by data point fusion when four-layer LIDAR detects adjacent obstacles in complex road environment,an adaptive distance threshold is proposed according to the characteristics of vehicle obstacle data,and an optimization method of adjacent obstacle detection is proposed combining with the distribution and change rule of adjacent obstacle data points.Experimental results show that this method can improve the accuracy of detection of adjacent obstacles in complex road environment...