Multi-Targets Detection And Location And Mapping Based On Millimeter-Wave Radar

The vehicle millimeter wave radar is one of the important sensors for unmanned driving.Compared with other sensors,the millimeter wave radar has the characteristics of all-weather,low price and convenient installation,and has broad application prospects.Due to the complicated road conditions,there are strong ground object clutter in the radar echo.How to stably track multiple targets in the clutter background is a problem that needs to be solved.At the same time,due to the detection performance of a single radar sensor,it has a certain time and Space limitation,multi-sensor fusion technology solves this problem.Through the multiple sensor information arranged on the car body,the detection probability of the target can be improved and the range of target tracking can be expanded.In unmanned driving,the positioning of the vehicle is an important issue.The accuracy of the positioning affects the decision of the vehicle.Since the radar echo data contains a large number of fixed target points in the environment,then these fixed target points are used.Positioning the vehicle while framing the environment is a problem worth studying.With the rapid development of unmanned driving,the use of vehicle-mounted millimeter-wave radar for environmental sensing and vehicle positioning has become a research hotspot.How to make full use of the characteristics of vehicle-mounted millimeter-wave radar,stable multi-target tracking for moving people and vehicles,vehicle positioning and environmental composition for fixed targets in the environment is a problem that needs to be solved.This paper combines the characteristics of millimeter-wave radar.To study the application of millimeter wave radar in environmental sensing and vehicle positioning,the main work includes the following aspects:First,the multi-target tracking problem in the case of single sensor is studied.In multi-target tracking,data association is an important step.Based on the interconnection of probabilistic data,the method of joint integration probability data association with the existence probability of track is studied.The algorithm models the existence problem of track as one.The Markov process,recursively calculates the existence probability of the track,so that the determination and deletion of the track are classified into a unified process.Aiming at the large starting error of the traditional track starting method,a track starting method based on normal velocity estimation is proposed.The method estimates the normal velocity of the target based on two adjacent track points,combined with the radial direction of the target.The velocity information estimates the target velocity,effectively reducing the initial error of the state of the track,and reducing the convergence time of the track state.Secondly,due to the limitation of time and space detection of a single sensor,the multi-sensor fusion method can effectively solve this problem.This paper studies the centralized multi-sensor fusion method,using a fusion center to arrange six around the body.The radar sensor data is processed uniformly,achieving 360-degree stable tracking of the target,which greatly improves the coverage and detection probability of the target tracking.Finally,based on the fixed target points in the radar data,the problem of vehicle positioning and composition is carried out.By detecting the lanes of the radar data,scanning matching is performed between the multiple frames to locate the position of the vehicle in the environment.By merging the information of the vehicle’s own inertial sensors,the environment is patterned to provide information for unmanned decision making.