| Pedestrian recognition based on automobile millimeter-wave radar has the capabilities to accurately identify pedestrians in the road environment,and effectively reduce the collision between vehicles and pedestrians,which is currently a research hotspot in the field of intelligent traffic.The traditional algorithm about computer vision and pattern recognition which uses a camera to obtain image information and recognizes pedestrians has achieved high detection accuracy,but it is commonly influenced by dim light and poor-condition climate.The automobile millimeter-wave radar has the working advantages of all-time and all-weather,and the micro-motion information of the target in radar echo is an important means to achieve the fine recognition of the target,which has become an important development direction in the technology of pedestrian recognition.Focusing on the micro-motion characteristics of moving targets in the echo of automobile millimeter-wave radar.under real road conditions,the research content includes moving target detection,analysis to micro-motion feature of pedestrian and pedestrian recognition while the radar remains stationary,and relevant theoretical analysis,method research,simulation verification,and outfield testing are carried out.The main contents are as follows:1.For the problem of matching difficulty and range-velocity coupling in multi-target detection,an algorithm for moving target detection based on the Frequency Modulated Continuous Wave(FMCW)radar in 77 GHz frequency is proposed by analyzing the waveform and frequency of the automobile millimeter-wave radar,which can solve the difficult problem of multi-target matching and realize the range-velocity decoupling of targets.2.Aiming at the problem that the micro-motion of the pedestrian is complex and the radar echo is difficult to observe,the Boulic's motion model of human is studied,and the echo model of pedestrian based on FMCW millimeter-wave radar is derived,which lays a theoretical foundation for the research of pedestrian recognition of the automobile millimeter-wave radar.3.To solve the problem that the time-frequency maps of pedestrian obtained by the traditional methods of time-frequency analysis are multi-component aliasing heavily,a method of sparse time-frequency analysis based on compressed sensing is proposed,which solves the problem of multi-component aliasing of time-frequency map and can more observe the micro-motion characteristics of time-frequency maps effectively.4.It is difficult to achieve pedestrian detection because the long-time coherent accumulation of FMCW radar leads to Doppler spread.In this thesis,the echo model of pedestrian based on FMCW millimeter-wave radar is applied,and the algorithm of pedestrian recognition based on Doppler compensation and nonlinear phase compensation is proposed,which can achieve multi-pedestrian recognition accurately.Through simulation experiments,the correctness of the echo model of pedestrian on FMCW millimeter-wave radar is verified,and the effectiveness of the method of sparse time-frequency is confirmed.Simultaneously,the test environment was built using the FMCW radar in 77 GHz frequency,and the test results show that the algorithm of pedestrian recognition can achieve multi-target detection and pedestrian recognition in the road environment effectively. |
PDF Full Text Request