一种低信噪比图像序列小目标实时检测算法  

An algorithm of low SNR small targets real-time detection in imagery

作　　者：陈尚锋[1] 肖山竹[1] 卢焕章[1] 

机构地区：[1]国防科学技术大学ATR国防重点实验室,湖南长沙410073

出　　处：《信号处理》2009年第4期601-606,共6页Journal of Signal Processing

基　　金："863"高技术计划基金资助课题(2004AA813031)

摘　　要：图像序列弱小运动目标的实时检测,是成像探测制导系统中的关键技术之一。提出一种基于动态规划与轨迹关联的目标实时检测算法。该方法利用指数衰减和运动方向加权动态规划积累目标能量,然后进行门限判决和对候选目标进行轨迹关联检测,从而剔除假目标,给出真目标信息。该算法结构简洁,易于进行软硬件分割实现,实时性能好。实验表明该算法能实现对深空背景中信噪比大于1.8、速度范围0～1像素/帧多个目标的有效检测。The real-time observation of dim small moving targets in image sequences is one of the key techniques in imaging detection and guiding system. This paper proposes a dim small moving target detection algorithm based on a developed dynamic programming （DP） and track association algorithm. Enhancements are achieved by performing exponential decay energy accumulation and directional weight operations precede to DP. The method is easy of software and hardware segmentation, excellent real time performance. Experiments have been conducted and it is shown that low SNR（ 〉 1.8） moving targets with arbitrary direction and maximum velocity of 1 pix/frame can be detected with high probability.

关 键 词：弱小目标检测 图像序列 动态规划 实时检测 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]