基于上下层MAS的SAR图像海洋分布目标检测  

Distributed targets detection based on top and bottom multi-Agent system

作　　者：胡希驰[1] 朱敏慧[1] 张永光[2] 王小青[1] 

机构地区：[1]中国科学院电子学研究所微波成像技术国家重点实验室,北京100190 [2]中国科学院系统科学研究所,北京100190

出　　处：《系统工程与电子技术》2009年第12期2830-2834,共5页Systems Engineering and Electronics

基　　金：国家自然科学基金(40706063)资助课题

摘　　要：针对SAR图像中的海洋分布目标,从人的认知过程出发提出了一种基于上下层多Agent系统(topbottom multi-Agent system,TBMAS)检测算法。该算法针对上下层Agent定义了不同的行为。下层Agent获得初步的视觉知觉,上层Agent加入联想分析等处理获得最终检测结果。算法从分布目标的局部统计特性考虑,不需要进行统计模型和参数的估计,利用局域直方图和Kullback-Leibler散度进行差异度分析。实验表明,对于多种分布目标均有较好的检测效果。To detect the sea distributed targets in SAR images, a top and bottom multi-agent system （TBMAS） using local histograms is proposed. According to the human recognizing progress, different behaviors of the top Agents and the bottom ones are defined respectively. Bottom Agents get the primary optical fea- tures, while by imagining and pre-linking behaviors top Agents obtain the final detection result. Considering the local statistics features of the distributed targets, local histograms and KL divergence are used to analyze the different measure without statistics models and parameter estimation. Experiments show that the method is quite effective for several different kinds of distributed targets.

关 键 词：SAR图像 AGENT 海洋分布目标 局域直方图 KL散度 

分 类 号：TN957.52[电子电信—信号与信息处理]