Radar High Resolution Range Profile Target Recognition Based On Frame Segmentation And HMM

High Resolution Range Profile(HRRP)is the amplitude of coherent summations of the complex time return from target in each range resolution cell,which represents the scattering intensity distribution of the target scatting centers onto the radar line-of-sight.Since it contains the target structure signatures,radar HRRP target recognition has received intensive attention from the radar automatic target recognition community.In this paper,we focus on the process of HRRP target recognition,and discuss in depth the elimination of attitude sensitivity of HRRP in the pre-processing stage,the method of template construction in training stage.and the improvement of Hidden Markov Model(HMM)classification algorithm in the classification recognition stage.The main contents are summarized as follows:(1)An adaptive frame segmentation method based on mutual information is proposed for the attitude sensitivity problem of HRRP.The method uses the mutual information as the similarity measure in the frame segmentation method,and iteratively calculates and merges the two frames with the largest mutual information.This method is more reasonable than the equalinterval frame segmentation under attitude disturbance and overcomes the unestimable problem of the number of frames and the manual dependence of threshold adjustment in the traditional adaptive frame segmentation.Through modeling and simulation experiments,it is proved that this method has better recognition effect than the traditional frame segmentation.(2)A template construction method based on the maximum information profile is proposed for the problem of construct the template library of HRRP training samples.In traditional template extraction,a lot of HRRP details are often lost in the process of generating mean range profile,resulting in limited classification performance.By calculating the amount of information of all training samples in the frame and using the maximum information profile as the frame template.This method can retain more feature information in frame and provide rich details for classifier,thus improving the performance of classification recognition.It is demonstrated through modeling and simulation experiments that the template extraction based on the maximum information profile can retain more physical features of the target and outperforms the template extraction method based on the mean range profile in terms of recognition rate.(3)This paper studies the statistical model representation technique for target recognition HRRP data,and proposes a common initial HMM construction method.Traditional HMM has the following limitations when describing HRRP data as a statistical model: The observation probability matrix with Gaussian distribution is not strong enough to describe nonlinear HRRP data;In the training stage,the initial HMM model is generated independently in each frame,which requires a large amount of calculation.Therefore,this paper proposes a method to establish a common initial HMM.When using HMM to describe HRRP data,this method learns all HRRP samples and establishes a common initial observation probability matrix,so that HMM can describe HRRP data more reasonably and reduce the amount of calculation in the training stage.Finally,the effectiveness of the proposed method is proved by modeling and simulation experiments.