Research On Target Detection Technology Of High Resolution Millimeter Wave Radar

Compared with infrared,laser,camera and other sensors,millimeter-wave radar is irreplaceable in some environments such as night,heavy fog,and heavy rain due to its unique characteristics of all-weather work.Using millimeter-wave radar to detect targets can get farther distance information than lidar.Traditional single-chip millimeter-wave radars cannot detect complex environments and form rich three-dimensional point cloud data due to their insufficient angular resolution.They have been only used as auxiliary sensors for vehicle-mounted radars.If single-chip millimeter-wave radars are cascaded together to form a cascaded high-resolution millimeter-wave radar,a multi-transmit and multi-receive antenna can be configured.Combined with MIMO theory,the angular resolution can be effectively improved,and the detection and perception of complex environments can be achieved,Get rich 3D point cloud data.Firstly,the time-frequency diagram of the millimeter wave radar waveform is introduced,combined with the MIMO theory,the principles of ranging,speed and angle measurement are derived,and the constant false alarm detection is also introduced.In this paper,MIMO uses time division multiplexing.Time division multiplexing will cause the angle measurement to be ambiguous and the maximum measurable speed is reduced.The specific causes of the two problems are analyzed and deduced in principle,and a solution is proposed,that is,phase compensation for angle measurement ambiguity,exceeding the maximum measurable speed target,and using the virtual antenna virtualized by MIMO to restore the Doppler bandwidth.Compare some differences between single-chip radar and four-chip cascade radar,such as beam width,angle measurement resolution,and point cloud imaging,etc.,and compare them through experiments,and then perform cluster analysis on the formed point cloud,mainly studying K-means clustering algorithm,DBSCAN clustering algorithm,combined with actual collection scenarios to compare data clustering indicators.Secondly,a detailed introduction to the principle and formula derivation of Kalman filter and extended Kalman filter were carried out respectively,and simulation experiments were carried out on the two filtering algorithms,and a coordinate system suitable for the two kinds of filtering was established.Perform theoretical analysis on the data association,tracking start and end,maneuver decision,tracking gate and other modules involved in multi-target tracking,and analyze the nearest neighbor data association algorithm,probabilistic data association algorithm,and joint probabilistic data association in the data association algorithm.The algorithm is researched,and the tracking difference between the four-chip cascade radar and the single-chip radar is compared and analyzed through the actual measurement data.Finally,the construction and debugging of the experimental platform are carried out,and the workflow of the hardware platform is introduced.MIMO radar parameter configuration is carried out through the LUA script language,point cloud imaging and multi-target tracking experimental results are analyzed on the platform,and point cloud imaging and multi-target tracking functions are realized.The results show that the high-resolution cascade radar can not only perform accurate point cloud imaging of near and far targets,but also accurately track and restore the trajectory of the target.