Research On Traffic Target Detection And Recognition Method Based On FMCW Radar

The data collection mechanism of traffic target detection and recognition technique can be mainly divided into two types: radar and video collection equipment.Video collection equipment cannot complete the task of data collection well if it’s in extreme environments such as strong light,darkness,and fog.In contrast,Frequency Modulated Continuous Wave radar has the advantages of stability,convenience,and wide application conditions.Based on FMCW radar,the following researches are carried out on the detection and recognition methods of traffic targets.(1)There is a large amount of clutter interference under the urban traffic background.For different interferences,the different detection methods of traffic targets are explored.In the fast and slow time domain,for static interference,dynamic disturbance and target parameters extraction,combined with the working principle and signal model of FMCW radar,the moving target indication algorithm,two-dimensional CFAR algorithm and moving target detection algorithm are explored and experiments are carried out.The simulation gives the most suitable detection model.(2)In order to solve the problem that the single-dimensional target feature cannot fully reflect the target characteristics,a multi-dimensional feature spectrum construction method based on target micro-Doppler features and target motion parameters is studied.Based on the results of radar signal processing,a Range-Doppler Map which reflects the instantaneous characteristics of the target can be obtained;based on the Range-Doppler Map,the target velocity information and time information can be coupled to obtain a Doppler-Time Map;coupling the target range information and time Information can get a Range-Time Map.Through the radar detection algorithm and the parameter-time composition method,the instantaneous micro-Doppler feature and the time-series micro-Doppler feature of the traffic target are obtained,which fully characterizes the target’s motion state information.(3)In view of the possible data anomalies,data imbalances,feature concealment and other issues that may exist in the radar raw data,the data preprocessing method is explored,and a lightweight convolutional neural network model is proposed to identify traffic targets.First,based on the method of constructing the target motion feature map,the motion feature data set was established from three dimensions.In order to solve the problems of data abnormality,data imbalance,and data feature concealment that may exist in the data set,processing methods such as sub-sampling,data normalization,and data enhancement are respectively proposed.At the same time,a method of image fusion along the later stepping direction of the convolution kernel is proposed to establish a multi-dimensional feature fusion data set,which keeps all the features of the target and simplifies the network structure.(4)Using the traffic target recognition model in this paper,the experimental test and result analysis are carried out.Analyze the impact of different motion feature data sets extracted according to the different motion parameters of the target on the accuracy of target recognition,and improve the initial network model of this article based on the Squeeze Net light convolutional neural network,and improve the model’s recognition accuracy.