Research On Multi-target Detection Algorithm Based On Information Fusion Of Lidar And Millimeter Wave Radar

Target detection based on multi-sensor information fusion technology plays an important role in the perception of autonomous driving environment.However,the following difficulties exist in the multi-target detection process based on information fusion of lidar and millimeter-wave radar:1)Lidar is easy to miss detection due to the occlusion of the target during the measurement process;2)How to accurately track the target according to the information collected by lidar and millimeter wave radar;3)How to reasonably associate and fuse the targets tracked by lidar and millimeter-wave radar to accurately obtain the status of the tracked targets.For this reason,the multi-target detection method based on the information fusion of lidar and millimeter-wave radar was studied.Firstly,the paper established the kinematics model of the autonomous vehicle based on the constant turn rate and velocity(CTRV)model,and used the rotation and translation matrix to unify the lidar and the millimeter-wave radar in space;according to the different sampling frequencies of the two sensors,the time unification is performed.Secondly,the statistical filtering algorithm was used to eliminate the useless information collected by the sensor due to noise and its own measurement accuracy,and the Random Sample Consensus(Ransac)algorithm is used to segment the ground point cloud;At the same time,in order to reduce the influence of the near-dense and far-sparse phenomenon in the point cloud on the clustering results,the Euclidean clustering algorithm based on distance partitioning clusters the objects on the ground.And compared with the fuzzy C-means clustering algorithm.Thirdly,the Hungarian matching algorithm based on Mahalanobis distance correlates the preprocessed data of the two sensors.Next,in order to reduce the computational complexity of the unscented Kalman filter algorithm in the information fusion process,a sensor information fusion algorithm is proposed based on the unscented Kalman filter and covariance intersection,and a distributed information fusion system was employed as a fusion structure.Finally,Matlab simulation software was used to build a virtual encironment for multi-target detection based on lidar and millimeter-wave radar to verify the feasibility of the algorithm in the contribution.Among them,the Generalized Optimal SubPattern Assignment(GOSPA)indexs were used to evaluate and analyze the detection accuracy of the algorithm.In addition,in order to verify the effectiveness of the algorithm in this paper in the real traffic environment,the public dataset Nuscenes is selected for verification.Simulation results show that the proposed algorithm based on unscented Kalman filter and covariance intersection information fusion can outperform Joint Probabilistic Data Association(JPDA)and Gaussian mixture probability hypothesis density(GM-PHD)algorithms for multi-target detection.Therefore,the unscented Kalman filter and covariance intersection information fusion algorithm provides a new idea and basis for the research of multi-target detection based on information fusion of lidar and millimeter wave radar.