高空间分辨率遥感影像的多智能体分割方法研究  

High Spatial Resolution Remote Sensing Image Segmentation Based on Multi-agent Theory

作　　者：赵贝[1] 钟燕飞[1] 张良培[1] 

机构地区：[1]武汉大学测绘遥感信息工程国家重点实验室,湖北武汉430079

出　　处：《测绘学报》2013年第1期108-115,122,共9页Acta Geodaetica et Cartographica Sinica

基　　金：国家973计划(2011CB707105);国家自然科学基金(40901213;40930532);全国优秀博士学位论文作者专项资金(201052);教育部新世纪优秀人才支持计划(NECT-10-0624);中央高校基本科研业务费专项资金(3103006)

摘　　要：提出一种基于多智能体的高空间分辨率遥感影像分割算法(high spatial resolution remote sensing imagesegmentation algorithm based on multi-agent theory,MARSS)。该方法在区域合并中结合了影像的光谱信息和形状信息,同时利用多智能体与图像环境交互性强,灵活性高,具有并行运算的优点,通过多个智能体控制不同区域的合并过程,能够使分割算法的全局合并控制更加优化。试验结果表明,该算法的分割效果要优于分形网络演化算法(FNEA)。A novel segmentation algorithm based on multi-agent theory,namely MARSS,to perform the task of segmentation for high spatial resolution remote sensing imagery is preposed.The algorithm combines spectral and shape information in region merging,employs a number of agents to control the merging procedure in different regions and may make the global merging control more optimal by utilizing the advantages of multi-agent system,such as strong interaction,high flexibility,and concurrently control.By comparison with fractal net evolution approach（FNEA）,one of common segmentation for high spatial resolution remote sensing imagery,experimental results show that the segmentation results of the proposed algorithm are more effective than FNEA.

关 键 词：高空间分辨率 分割 多智能体 遥感 全局控制 

分 类 号：P237[天文地球—摄影测量与遥感]