基于数据场和密度聚类的高分辨率影像居民区提取  

Residential area extraction for high resolution remote sensing image based on data field and density clustering

作　　者：岳梦雪 秦昆[1,2] 张恩兵 张晔[1] 曾诚[1] 

机构地区：[1]武汉大学遥感信息工程学院,武汉430079 [2]地球空间信息技术协同创新中心,武汉430079

出　　处：《国土资源遥感》2017年第3期92-97,共6页Remote Sensing for Land & Resources

基　　金：国家重点基础研究发展计划(973)项目"高分辨率遥感影像的目标特征描述与数学建模"(编号:2012CB719903);重庆市国土房管局科技计划项目"基于图像识别技术的国家高分辨率遥感数据分析应用方法研究"(编号:CQGT-KJ-2014032)共同资助

摘　　要：数据场通过模拟物理场中对象间的相互作用,来描述数据对象间的相互作用关系。数据场中的势值高低反映对象间相关程度,故在遥感影像中可用数据场来刻画像元间的空间相关性特征。提出了一种基于数据场和密度聚类的高分辨率居民区有效提取的方法。首先,利用数据场计算遥感影像的势值特征图像;然后,对势值图像进行分水岭分割,提取分割所得对象块的形心;最后,对形心进行基于密度的聚类,从而实现居民区提取。实验结果表明,基于此方法进行高分辨率遥感影像的居民区提取相对于传统方法具有更好的鲁棒性和高效性。Data field can describe the correlation between data objects,and is a simulation of interaction between particles in physical field. Potential value of a data object in data field can effectively represent the spatial interactions of its neighborhoods,and it can do so for pixels in high resolution remote-sensing image. In this paper,the authors propose a method for residential area extraction from high resolution remote-sensing image using data field and density clustering. The major steps are as follows: the generating of a high resolution remote-sensing image data field; the calculation if potential value for each pixel in this field to obtain a new feature image;the segmentation of the feature image via watershed segmentation,and the calculation of centroids of segmentation results; the clustering of all the centroids into different clusters based on the density,with the extracted residential area composed of target clusters. Compared with existing relative methods of residential areas extraction for high resolution remote-sensing images,the experimental results suggest that the presented method is robust and efficient.

关 键 词：数据场 空间相关性 密度聚类 高分辨率遥感影像 居民区提取 

分 类 号：TP79[自动化与计算机技术—检测技术与自动化装置]