阵列相机图像邻近目标超分辨方法  

Closely Spaced Objects Super-resolution Method Using Array Camera Images

作　　者：安成锦 杨俊刚 梁政宇 陈芊羽 曾瑶源 安玮[1] AN Chengjin;YANG Jungang;LIANG Zhengyu;CHEN Qianyu;ZENG Yaoyuan;AN Wei(College of Electronic Science and Technology,National Unibersity of Defense Technology,Changsha 410073,China)

机构地区：[1]国防科技大学电子科学学院,长沙410073

出　　处：《电子与信息学报》2023年第11期4050-4059,共10页Journal of Electronics & Information Technology

基　　金：国家自然科学基金(61921001);“卫星信息智能处理与应用技术”实验室自主科研基金(2022-ZZKY-JJ-14-01)。

摘　　要：空天威胁目标通常距离成像系统较远,导致其在图像中信噪比低、尺寸小,即呈现为弱小目标。由于系统分辨率限制,当目标以密集目标群出现时,往往在图像中形成未分辨目标簇,对目标发现、跟踪、识别等带来挑战。阵列相机可以提供多个视角的互补观测信息,采用融合阵列相机图像的超分辨技术,可有效提升弱小目标分辨能力,为分辨密集多目标提供技术途径。该文分析了空间邻近目标与阵列相机之间的几何关系,并提出一种基于阵列相机图像稀疏重建的邻近目标超分辨率方法。利用空间邻近目标在像平面上稀疏性先验假设和阵列相机多视图之间关于目标的投影约束,仿真实验结果表明所提方法能够有效分辨空间邻近目标,实现对空间邻近目标位置和数量的准确估计。The aerial targets are usually far from the imaging system,making the imaged results have weak radiation intensity and limited imaging area.Especially when aerial target groups are distributed in a dense form,make further the imaged results have overlapping projection of such dense targets,and limit the performance of subsequent detection,track,and identification tasks.The array camera imaging system can provide complementary information about the target from multiple views and make effectively up for the deficiency of a single camera in detecting the resolution of nearby aerial targets.In this paper,the geometric relationship between nearby aerial targets and array cameras is studied and a super-resolution method for nearby targets based on sparse reconstruction of array camera images is proposed.Thanks to the prior assumption of sparsity of nearby aerial targets on the image plane and the transfer constraints between multiple views of array cameras regarding the target,relevant simulation experiments show that the proposed method can well super-resolve the obtained images of nearby targets and estimate effectively the position and number of nearby aerial targets.

关 键 词：阵列相机 邻近目标 超分辨率 稀疏重构 

分 类 号：TN911.73[电子电信—通信与信息系统]