A Generative Approach for Incomplete High Resolution Range Profile Sequence Recognition

Abstract

The High-Resolution Range Profile (HRRP) sequence recognition has attracted great concern in the field of Radar Automatic Target Recognition (RATR). In this paper, a novel HRRP sequence recognition method based on Restricted Boltzmann Machine (RBM) is proposed for the poor performance of the traditional methods on incomplete data and noise corrupted data. Two main procedures are included in our method, which are estimation and recognition. Firstly, infinite Restricted Boltzmann Machine, as a generative model, can estimate the missing frame HRRP signal by learning the features of several adjacent frames of the missing HRRP data. Additionally, the original incomplete sequence combined with the estimated HRRP constitutes a new complete sequence, which is utilized to complete the recognition task. Therefore, our method provides a novel idea for the identification of incomplete data and has strong robustness to noise. Experiment results on the Moving and Stationary Target Acquisition and Recognition (MSTAR) dataset show that our proposed model outperforms other traditional methods, indicating that our method is suitable for incomplete data and noise corrupted data.