基于轮廓小波变换的航空影像建筑物识别  

Recognition of Buildings from Airborne Remote Sensing Image Based on Contour Wavelet Transform

作　　者：吴秀芸[1] 周华 王海江[1] 李艳[3] 

机构地区：[1]宁波市测绘设计研究院,浙江宁波315042 [2]宁波弘泰水利信息科技有限公司,浙江宁波315042 [3]南京大学国际地球系统科学研究所,江苏南京210093

出　　处：《地理信息世界》2014年第6期114-117,共4页Geomatics World

摘　　要：在对航空影像中的建筑物进行识别提取时,建筑物顶部的轮廓信息是一个重要的判断依据。基于航空影像建筑物个数繁多、形状复杂,且存在较多的干扰信息,提出一种新颖有效的建筑物识别方法:首先,利用改进标记分水岭算法提取建筑物区域。然后,对每个分割得到的建筑物区域,提取其轮廓,对轮廓进行基于平稳小波变换的仿射不变量计算,并构造建筑物模型数据库,利用相关系数实现了建筑物的有效识别。In the process of buildings recognition, the contour information of rooftops is an important judgment. In the paper, a novel and effective identification method is proposed for aerial images which contain many buildings of complex shapes and much noise. Firstly, the improved marker-controlled watershed algorithm is used to obtain building regions. Then, in each region segmented, the contours are extracted to calculate Affine Invariants based on Stationary Wavelet transform, building model database is also constructed. Finally, Correlation is used to effectively recognize buildings.

关 键 词：航空影像 改进标记分水岭 平稳小波 仿射不变量 相关系数 

分 类 号：P2[天文地球—测绘科学与技术] TP392[自动化与计算机技术—计算机应用技术]