Research On Person Location Algorithm Based On Millimeter-Wave Radar

The millimeter-wave radar positioning technology realizes the position estimation of the person based on the signal echo reflected by the human body.Because the millimeter-wave signal has strong environmental adaptability and certain penetration ability,high ranging accuracy,relatively low cost,and does not directly collect human sensitive biometric information,it has a certain privacy protection effect.Therefore,it can be applied to scenarios such as unmanned mobile platform navigation,smart home,and building monitoring.These application scenarios usually need to complete the tasks of estimating the current position of the person and predicting the future trajectory of the person.However,the low angular measurement resolution of millimeter-wave radar limits its ability to locate the current position of a person,and the pedestrian’s trajectory is easily affected by other pedestrians and changes frequently,which poses a challenge to the prediction of the millimeter-wave radar’s pedestrian trajectory.In order to solve the above problems,this thesis has done the following research:(1)In order to improve the positioning accuracy of millimeter-wave radar for personnel,this thesis proposes a millimeter-wave radar fusion localization algorithm based on cross-modal learning--mm Fusion.Because millimeter-wave radar has a strong range measurement capability and visual positioning has a strong angular measurement capability,mm Fusion first designed a cross-modal personnel position prediction module,using neural networks to model the relationship between visual positioning results and radar signal data.It can improve the angular positioning capability of millimeter-wave radar.Then use the signal processing chain personnel position estimation module to process the radar signal to give full play to the ranging capability of the millimeter-wave radar.Finally,mm Fusion integrates the two positioning modules of cross-modal personnel position prediction and signal processing chain personnel position estimation to improve the positioning accuracy of millimeter-wave radar.The experimental comparison with other millimeter-wave radar positioning methods shows that mm Fusion effectively reduces the positioning error and improves the accuracy of personnel detection.(2)In order to improve the accuracy of pedestrian trajectory prediction by millimeter-wave radar,this thesis proposes a Transformer-based Trajectories Prediction Algorithm(TTPA).TTPA first used the bipartite graph-based historical trajectory tracking module to complete pedestrian trajectory tracking.Then a Transformer-based pedestrian trajectory prediction model is designed to complete the pedestrian trajectory prediction.The prediction model uses the adjacent historical trajectory encoder and the future trajectory encoder to deal with the pedestrian trajectory changes caused by others.The experimental comparison with other pedestrian trajectory prediction algorithms shows that TTPA effectively reduces the average displacement error and final displacement error of pedestrian trajectory prediction.