基于数据场和密度聚类的高分辨率影像居民区提取
发布时间：2018-11-01 15:36
【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。
[Abstract]:The data field describes the interaction between data objects by simulating the interaction between objects in the physical field. The potential values in the data field reflect the degree of correlation between objects, so the spatial correlation characteristics between pixels can be described in remote sensing images. A high resolution neighborhood extraction method based on data field and density clustering is proposed. Firstly, the potential feature image of remote sensing image is calculated by using the data field; then, the potential value image is segmented by watershed, and the centroid of the object block is extracted. Finally, the centroid is clustered based on density to realize the extraction of residential area. The experimental results show that the proposed method is more robust and efficient than the traditional method in the extraction of residential areas from high-resolution remote sensing images.
【作者单位】： 武汉大学遥感信息工程学院;地球空间信息技术协同创新中心;
【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
【分类号】：TP751
本文编号：2304354
[Abstract]:The data field describes the interaction between data objects by simulating the interaction between objects in the physical field. The potential values in the data field reflect the degree of correlation between objects, so the spatial correlation characteristics between pixels can be described in remote sensing images. A high resolution neighborhood extraction method based on data field and density clustering is proposed. Firstly, the potential feature image of remote sensing image is calculated by using the data field; then, the potential value image is segmented by watershed, and the centroid of the object block is extracted. Finally, the centroid is clustered based on density to realize the extraction of residential area. The experimental results show that the proposed method is more robust and efficient than the traditional method in the extraction of residential areas from high-resolution remote sensing images.
【作者单位】： 武汉大学遥感信息工程学院;地球空间信息技术协同创新中心;
【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
【分类号】：TP751
【相似文献】
相关期刊论文 前1条
1 马建文,刘素红,马超飞;遥感多维空间数据场特征的角度分析与应用[J];遥感学报;2001年01期
，本文编号：2304354
本文链接：https://www.wllwen.com/guanlilunwen/gongchengguanli/2304354.html
