微波成像中的參數(shù)化稀疏表征模型

李剛 張顥 孟華東 劉一民 王希勤

摘 要：稀疏表征的任務(wù)是找到一個(gè)基信號(hào)矩陣，在雷達(dá)回波數(shù)據(jù)域和稀疏域之間構(gòu)建一個(gè)線性映射。經(jīng)典稀疏表征模型中，基信號(hào)矩陣是預(yù)先設(shè)定的，例如：傅里葉矩陣、小波矩陣等等，而且在稀疏求解過程中是固定不變的。然而，雷達(dá)目標(biāo)往往存在非合作運(yùn)動(dòng)，這將給雷達(dá)回波帶來未知的距離徙動(dòng)和頻率調(diào)制，導(dǎo)致傳統(tǒng)基矩陣無法實(shí)現(xiàn)非合作目標(biāo)回波信號(hào)的稀疏表征。為解決這一難題，提出了參數(shù)化稀疏表征模型，構(gòu)建了以目標(biāo)特征狀態(tài)為參數(shù)的基信號(hào)矩陣，并實(shí)現(xiàn)了目標(biāo)運(yùn)動(dòng)狀態(tài)估計(jì)與稀疏恢復(fù)的聯(lián)合求解。仿真和實(shí)測(cè)雷達(dá)數(shù)據(jù)實(shí)驗(yàn)表明，參數(shù)化稀疏表征模型能夠有效地提高雷達(dá)圖像質(zhì)量。

關(guān)鍵詞：壓縮感知 雷達(dá)成像 稀疏表征 字典學(xué)習(xí)

Abstract：The goal of sparse representation is to find a dictionary matrix that maps radar signals onto a sparse domain.In traditional models of sparse representation，the dictionary is pre-designed and fixed during the solution process.The popular dictionaries include Fourier and Wavelet matrices.However，the non-cooperative motion of the target causes unknown range migration and frequency modulation. Therefore，traditional dictionaries cannot ensure the sparse representation of the echo from a non-cooperative target.To solve this problem，we propose parametric sparse representation model，create the dictionary related to target motion status parameters，and simultaneously achieve the sparse representation and the parameter estimation.Simulations and experiments on real radar data show that parametric sparse representation is helpful to improve the quality of radar images.

Key Words：Compressed sensing；Radar imaging；Sparse representation；Dictionary learning

閱讀全文鏈接（需實(shí)名注冊(cè)）：http：//www.nstrs.cn/xiangxiBG.aspx？id=51796&flag=1