HRRP Target Recognition And Classification Based On Millimeter Wave Radar

The external environment has little impact on millimeter wave radar so it can work stably under various conditions.Millimeter wave radar is widely used in certain traffic management,smart city and other scenarios,and the millimeter wave radar is also needed to identify the type of vehicle in intersection monitoring.Following the capabilities of millimeter-wave radar,the focus is on the High Resolution Range Profile(HRRP),the characteristics of the current scene are analyzed,and the recognition and classification method of vehicle targets in the current scene is designed,and the experimental simulation is carried out to verify effect.The work of this thesis is mainly divided into the following two parts:The first part studies the radar scatter point model,analyzes the sensitivity and solution of HRRP.Then it deeply studied the shortcomings of traditional algorithms.CNN-based recognition methods are dedicated to extracting local structural features in HRRP,but ignore the temporal correlation of samples.The RNN network lacks long-term dependence and has poor performance.This paper proposes the AC-LSTM algorithm,which uses one-dimensional CNN to extract local features in HRRP,and introduces a channel adjustment module to improve the model’s ability to recognize features of different channels and improve model performance.The extracted features are extracted with the bidirectional LSTM model for time series feature extraction,which can effectively improve the long-term dependence problem while making full use of the HRRP time series correlation.The attention mechanism is introduced to further improve the recognition accuracy.This method has strong interference performance against time shift and can work in a low signal-to-noise ratio environment.Simulation experiments using the MSTAR database also proved the performance of this method.The second part analyzes the problem of poor recognition results of measured data.In the actual scene,the environment of the vehicle target is complex,and ground clutter and roadside obstacles will interfere with the target echo and affect the recognition algorithm.Therefore,it is necessary to obtain the signal of the desired target through the target detection algorithm.However,due to the high range resolution of millimeter-wave radar,the target is often in multiple range units.If the traditional CFAR algorithm is used for target detection,the signal belonging to the target will be screened out,and a large amount of target information will be lost.This paper proposes a CFAR algorithm based on energy enhancement,which considers the energy of multiple points around it when detecting a target point,effectively detecting the target in the millimeter wave radar echo,and improving the model recognition ability in the actual scene.The simulation experiment using the measured database also proves the performance of this method.