【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。更多还原

【Abstract】 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 features,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.更多还原