基于图像域信息认知的弹载前视成像研究

For the missile-borne based looking-forward imaging system, when the moving direction of the platform is parallel to the beam pointing, no synthetic aperture is availabe for radar imaging, which restricts the development of radar imaging technology in missile-borne imaging system. for the looking-forward target, only the limited real array aperture can be utilized to get the scene information. However, due to the limited aperture in missile-bore system, there is not enough space to arrange more array antennas, and with the maneuverability of the platform, the stationarity of the data is destroyed, many super-resolution algorithms can not be combined to improve the image quality, as well as the image information. In this fund,we are proposed to introduce the cognitive feedback from the scene information in image domain into the process of the looking-forward imaging procedure, which make the information characteristic of the scene in image domain feedback to the imaging data domain, and construct a close-loop system from the physical modeling to cognitive in imaging domain, and finally achieve a new high resolution imaging system without modify any front-end structure. This cognitive imaging mode will break the classical imaging ideology, that only with large bandwidth and big array aperture, can high-resolution image quality be obtained, and imaging with high quality can be gotten via small aperture, it breaks the idealogy that no image can be availabe with small array antennas, and presents a novel methodology for missile-borne based looking-forward imaging system.

对于弹载前视成像系统来说，当波束指向与运动方向平行时无法形成虚拟孔径，导致无法采用合成孔径雷达进行成像，从而制约了雷达成像技术在弹载前视成像系统中的应用与发展。即对于前视目标来说，只能通过实孔径阵列获取场景信息。由于弹载设备口径有限，无法安置过多的阵列天线来获取场景的空间信息，加之实际载体平台运动的机动性，破坏了数据的平稳性，致使无法使用超分辨算法来提升成像质量，进而无法获得场景的成像信息。本项目拟在前视成像过程中引入图像域场景信息的认知反馈，即将场景信息在图像域的信息特征反馈到成像数据域，构成物理建模到图像域信息认知的闭环系统，从而实现不改变系统前端结构的高分辨成像。该认知成像模式将打破传统成像思维，即只有采用大带宽、大阵列孔径才能实现高分辨率成像的思想束缚，实现小孔径下高分辨率雷达成像，打破传统小孔径阵列雷达无法成像的思想，为弹载前视成像雷达的设计提供一种新思路。

本课题针对前视成像研究中遇到的成像效率，成像质量等问题展开了深入的研究，在成像建模中，通过引入可分离观测模型及可分离稀疏表示模型，大大简化了成像建模，提高了成像效率，与此同时，通过将图像域的全变差变化引入到成像建模中，提升了前视成像的分辨率，通过几年的研究，课题组在空域成像建模，基于稀疏表示的成像质量提升及成像效率提升等方面获得了诸多研究成果，发表论文多篇，申请并授权专利多项，本课题所开展的工作已经经过GPU并行平台论证，本课题方案的提出给后续前视成像的工程化研究提供了有力的技术保障和理论基础。