Research On Radar Imaging Method Based On Signal Sparse Representation

Radar imaging achieves the inversion of target scattering characteristics using echo data,which plays an important role in military application.In order to obtain accurate radar target characteristics,high resolution radar imaging techniques,including ISAR radar imaging and three-dimensional imaging,have been extensively studied.Signal sparse representation is an emerging signal analysis method.Based on the dictionary matrix,sparse representation of the signal is realized by observation data.The sparse nature of the signal is deeply dig,providing many novel solutions for radar imaging.The purpose of the paper is to improve the radar imaging quality and lay a foundation for target recognition.Based on the signal sparse representation analysis,multi-radar data fusion technology and compression sensing technology are used to accomplish the purpose.This thesis is mainly divided into following parts:The first part introduces the background and the basic theory of high resolution imaging,including the ISAR imaging,3d radar imaging,signal sparse representation theory,scattering center modeling compressing sensing theory and so on.The second part studies imaging technique based on data fusion technology in GTD model,analyzing multiband radar data fusion imaging and multiband multi-angle radar data fusion imaging.Building the signal sparse representation model for 1d electromagnetic scattering model and 2d electromagnetic scattering model.With known sub-band sub aperture radar echo,using BCS and expansion-compression variance component method(ExCoV)method to obtain the solution of the sparse signal and accomplish the process of radar data fusion imaging.The influence of different reconstruction methods on imaging quality under different noises is analyzed,and radar fusion high resolution imaging is realized in low SNR.In the third part,three-dimensional radar high resolution imaging technique based on tensor method is studied.First,the basic principle of 3d imaging is analyzed,while a 3d radar signal sparse representation model is established.Then traditional vector-based compression sensing method is used to improve the quality of three-dimensional radar imaging.For the fact that traditional vector-based CS method imposes a huge memory storage and computational burden,tensor multi-dimensional CS method MD-NSLO is studied by processing 3d radar echo data directly,which decreases the space complexity.MD-CS method provides a solution for the realization of 3d radar imaging in general computer.