Research On Key Technologies Of Millimeter Wave Detection For Multi-targets In Short-range

There are always many potential targets in the short-range and semi-open space with poor visibility,such as logistics parks,industrial plants,fire scenes,etc.And it is necessary for the vehicles working in these scenses to have the active safety ability of hazard perception and identification.As one of the mainstream vehicle sensors,frequency modulated continuous wave(FMCW)mmwave radar,which can work at all-weather and all-time with high resolution,is appropriate to the above scenes.However,the existing mmwave detection technology faces the problems of many interferences points and time-consuming in the short-range spaces with multi-targets.And this thesis focuses on the signal processing technologies of FMCW mmwave radar to further improve the performance of target detection,and the main contents of it are as follows:(1)The working principle of FMCW mmwave radar based on sawtooth modulation is analyzed as well as the signal processing algorithms,including FFT,constant false-alarm rate(CFAR)algorithm,signal incoherent accumulation,points cloud clustering,etc.And then,the CFAR algorithm and points cloud clustering are selected as the breakthrough points to improve the reliability of radar target detection.(2)The multi-targets detection effects of the mean level CFAR and the Order Statistics(OS)CFAR detection models in the environments of uniform clutter and clutter edge are simulated and analyzed.Finally,an improved fast OSCA-CFAR algorithm is proposed,and the processing efficiency of the algorithm is improved and a good effect for multi-targets detection is achieved in short range environment by applying the improved fast OS-CFAR based on the partial insertion sorting and the multi-cells mean value as detection threshold,and the improved CA-CFAR based on the normalized weighted matrix multiplication in the range dimension and doppler dimension of the RDM.(3)Aiming at the points cloud clustering,this thesis discusses the clustering algorithms commonly used,including k-means,DBSCAN,and focuses on the clustering algorithm of fast search and find of density peaks(CFSFDP),and analyzes its limitations and some improvements made by predecessors to CFSFDP.Finally,an improved adaptive CFSFDP algorithm is proposed to realize the self-determination of the cut-off distance d_c and the clustering centers.(4)The effectivenesses of the above improved algorithms for multi-targets detection in short range are verified based on TI’s AWR1642 mmwave radar.