目前,传统雷达成像方法的发展日渐完善,但在前视成像场景下,雷达难以获取方位向上的多普勒信息,从而限制了其方位向分辨率.为了解决这一问题,国内提出了微波关联成像方法.微波关联成像方法利用关联成像原理进行雷达成像,无需利用目标的多普勒信息即可实现高分辨率成像.这一新型雷达成像方法突破了传统雷达成像方法中受限于雷达孔径的分辨率,具有极高的前视成像发展潜力.目前,国内外对微波关联成像的研究主要集中在产生随机波前、解决模型失配问题和研制超材料孔径等方面,但对关键的关联过程的优化主要集中在压缩感知和深度学习方面,而在伪逆算法方面的研究相对较少.因此,为了进一步完善微波关联成像体系,本文提出了一种新的针对伪逆算法优化的微波关联前视成像方法.本文结合截断奇异值分解(Truncated Singular Value Decomposition,TSVD)处理和吉洪诺夫正则化(Tikhonov)提出了奇异值分解和吉洪诺夫正则化的联合处理方法(TSVD-Tikhonov,TSVDT),通过TSVDT方法对时空随机辐射阵进行处理,然后进行压缩关联成像.同时,本文比较了广义交叉验证(Generalized Cross-Validation,GCV)和L曲线法,并证明了在微波关联成像方法中,利用GCV法选择截断参数的运算耗时更短且更稳定.最后,利用微波暗室实验验证了该方法在低信噪比条件下提高了成像的抗干扰能力,并且仍能保持较快的运算速度.
Method of Microwave Correlation Forward View Imaging Based on TSVDT
Recently,traditional radar imaging methods have been considerably improved.However,in forward-looking imaging scenarios,radar faces difficulties in obtaining azimuthal Doppler information,thereby limiting its azimuthal resolution.To address this issue,this study proposes a microwave correlation imaging method.The microwave correla-tion imaging method utilizes the principle of correlation imaging to achieve high-resolution imaging without relying on target Doppler information.This novel radar imaging method overcomes the resolution limitations imposed by radar ap-ertures in traditional radar imaging methods and holds significant potential for forward-looking imaging.Currently,re-search on microwave correlation imaging,both domestic and international,primarily focuses on generating random wavefronts,solving model mismatch problems,and developing metamaterial apertures.However,the optimization of the crucial correlation process primarily focuses on compressive sensing and deep learning,with relatively limited re-search on pseudo-inverse algorithms.Therefore,to further improve the microwave correlation imaging system,this study proposes a novel microwave correlation forward-looking imaging method that optimizes the pseudo-inverse algo-rithm.This study combines truncated singular value decomposition(TSVD)processing and Tikhonov regularization to propose a joint processing method of SVD and Tikhonov regularization(TSVD-Tikhonov,TSVDT).The TSVDT method involves processing the spatio-temporal random radiation array and then performing compressed correlation im-aging.Furthermore,this paper compares generalized cross-validation(GCV)and L-curve methods and demonstrates that in microwave correlation imaging,using the GCV method for truncation parameter selection results in shorter com-putation time and greater stability.Finally,the proposed method was validated through microwave anechoic chamber ex-periments,which demonstrated its improved anti-interference capability in low signal-to-noise ratio conditions while maintaining high computational speed.
microwave coincidenceanterior vision imagingrandom FMtruncated singular value decompositionpseudoinverse algorithm
NETL
NSTL
万方数据
