论文部分内容阅读
基于像素模糊C均值算法(FCM)及其改进算法难以解决高分辨率遥感影像中地物目标光谱测度相似性减弱和几何噪声增大带来的分割难题,提出一种基于区域的FCM算法.该方法利用Voronoi几何划分将影像域划分为子区域,并用子区域拟合地物目标的几何形状.在此基础上,定义区域FCM目标函数,通过迭代最小化该目标函数实现高分辨率遥感影像分割.实验结果表明,与基于像素的FCM和增强FCM方法相比,所提出方法可以更加精确地实现高分辨率遥感影像分割.
Based on the pixel-based fuzzy C-means algorithm (FCM) and its improved algorithm, it is difficult to solve the segmentation problem caused by the reduced similarity of the target spectral measure in high-resolution remote sensing images and the increase of geometric noise. A region-based FCM algorithm is proposed. The method divides the image domain into sub-regions using Voronoi’s geometric partition, and uses the sub-region to fit the geometric shape of the object.On this basis, the FCM objective function is defined and the high-resolution remote sensing image segmentation is achieved by minimizing the objective function by iteration The experimental results show that compared with pixel based FCM and enhanced FCM, the proposed method can achieve high resolution remote sensing image segmentation more accurately.