Research On Detection Method Of Human Targets Of Millimeter-Wave FMCW MIMO Radar

The human target detection radar obtains the spatial position information of the human target by emitting electromagnetic waves on the detection scene,and then processing the echo signal such as clutter suppression,target detection and imaging.Human-computer interaction and other fields have broad application prospects and important application value.Compared with the existing low-band radar systems,millimeter-wave FMCW MIMO radar has the characteristics of small size and compact antenna,so the detection requirements and difficulties are different from those of low-band radar.On the one hand,the constant false alarm(CFAR)detection method is often used in the existing literature.Due to the small radar cross-sectional area(RCS)of the human body and the low signal-to-noise ratio(SNR)of the echo,the detection performance will deteriorate;On the one hand,the current millimeter wave radar often uses beamforming(BF)and Capon methods to estimate the angle,and the angular resolution is not high,and the high-resolution angle estimation method will bring about the problem of increased computational complexity.Focusing on the above problems,this paper will focus on the methods of improving SNR,high-resolution two-dimensional(2D)angle estimation methods and their rapid implementation,and the construction of radar systems.The specific work arrangement of this paper is as follows:1.Aiming at the problem of low echo SNR in the detection,the performance of target detection is degraded,and a joint time-dimensional and spatial-dimensional non-coherent integration(NCI)method is proposed to improve the performance of target detection.The method first performs windowed fast Fourier transform(FFT)and moving target display(MTI)processing on the beat signal to obtain the range spectrum and suppress static environment clutter,and then transmits multiple FMCW and MIMO radar synthesis based on FMCW radar The characteristics of the virtual array are combined with multiple echo signals in the time dimension and multiple virtual array echo signals in the space dimension to perform incoherent integration to improve the SNR of the echoes.Finally,the accumulated results are processed by logarithmic detection to reduce noise.The influence of the wave,the processing results are sent to the unit average selection small constant false alarm(CASO-CFAR)detector to detect the distance unit where the target is located.Simulation and experimental results show that the method improves the echo SNR and further improves the performance of target detection.2.Aiming at the problem of low resolution of 2D angle estimation methods in existing detection methods,a two-dimensional asymptotic minimum variance sparse iterative(2D SAMV)high-resolution angle estimation method is proposed.The method firstly extracts the echo signals in the array dimension and slow time dimension according to the detected distance information,constructs the multi-shot receiving data of the virtual array,and then uses the sparse characteristics of the target to divide the dense grid in the 2D angle to construct The over-complete steering vector matrix is then used to establish the iterative formula of grid power and noise power according to the asymptotic minimum variance(AMV)criterion,and then the azimuth-elevation power spectrum on the range unit is obtained,and finally 2D CFAR detection is performed on it.And according to the detection results,the 3D point cloud image of the target is obtained.The simulation and experimental results show that the 2D SAMV method improves the angular resolution compared with the Capon,MUSIC and IAA methods in the existing detection methods.3.The above-mentioned 2D SAMV method needs to further refine the imaging grid in order to obtain high-resolution performance,and the refinement of the grid will result in more grid power to be estimated,which in turn leads to an excessively long time for estimating the 2D angle spectrum.Aiming at this problem,a fast 2D SAMV angle estimation method based on adaptive grid evolution is proposed.The method first divides the coarse grid on the 2D angle of the distance element,then uses 2D SAMV to estimate the power of the coarse grid,and uses 2D CFAR for detection to determine the position of the coarse grid where the target exists,and then uses the detected coarse grid to estimate the power of the grid.The unit is refined and split at a fixed scale,and continues to use 2D SAMV to estimate the refined mesh,coupled with 2D CFAR detection,and finally obtain the 3D point cloud image of the target according to the detection result.Simulation and experimental results show that this method reduces the computation time of 2D angle estimation while ensuring high angular resolution.4.Build a millimeter-wave radar human target detection system for generating 3D point clouds and people counting.The system uses Texas Instruments(TI)millimeter-wave radar sensor IWR6843 ISK,data acquisition card DCA1000 EVM and PC to build the hardware platform of the system,and uses MATLAB to develop user graphical interface(GUI)software.After the system is powered on,first the PC controls the mm Wave Studio software to download the working parameters to the IWR6843 ISK through the GUI,the radar starts to work,and then the DCA1000 collects the radar data uploaded by the radar and transmits it to the designated position in the PC,and then the binary radar data is in the It is parsed,reorganized and processed on the PC by combining the above methods,and the3 D point cloud of the target is generated from this.Finally,the point cloud is processed by the DBSCAN clustering algorithm to obtain the result of people counting.show.The feasibility of the system for generating 3D point clouds of human objects and people counting in different scenarios is verified by experiments.