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.