摘要
针对高分辨率遥感影像中传统超像素分割方法存在过分割和边缘分割一致性的问题,选取五景高分辨率遥感影像,提出了一种融合光谱差异的空间约束模糊聚类的熵率超像素分割方法。首先,采用熵率超像素进行过分割,生成超像素过分割区域;然后,分析影像地物的空间信息,对比区域相似性;最后,采用空间约束模糊聚类和光谱差异进行区域合并,获取最终分割影像。定性和定量分析结果表明,该方法改善了过分割问题,能有效提高分割精度,使得分割影像的地物边缘一致性较优。
In view of the high spatial resolution remote sensing image,the traditional superpixel segmentation method has the problem of over-segmentation and edge segmentation consistency.In order to solve this problem,this paper selects five high-resolution remote sensing images and proposes an entropy rate superpixel segmentation method based on spatially constrained fuzzy clustering fusion of spectral differences.Firstly,the entropy rate super pixel is used for over-segmentation to generate regional super pixels.Then,by analyzing the spatial information of the image features,the regional similarity is compared.Finally,the fuzzy clustering model of spatial constraints and spectral differences are combined to obtain the final segmented image.It is analyzed by qualitative and quantitative methods.Experimental results show that this method reduces the problem of over-segmentation,effectively improves the accuracy of segmentation,and makes the edge consistency of the segmented images better.
作者
陈佳旺
王征强
于庆和
CHEN Jiawang;WANG Zhengqiang;YU Qinghe(College of Geomatics,Xi’an University of Science and Technology,Xi’an 710000,China;BaojiSurvey and Mapping Institute,Baoji,Shaanxi 721000,China;Heilongjiang Academy ofForestry Design and Research,Haerbin 150080,China)
出处
《遥感信息》
CSCD
北大核心
2021年第5期74-80,共7页
Remote Sensing Information
基金
国家自然科学基金项目(41501571)。
关键词
过分割
熵率超像素
空间约束模糊聚类
光谱差异
边缘一致性
over-segmentation
entropy rate superpixel
spatially constrained fuzzy clustering
spectral difference
edge consistency
