CAJ | 학술논문

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。
재대항공영상중적건축물진행식별제취시，건축물정부적륜곽신식시일개중요적판단의거。기우항공영상건축물개수번다、형상복잡，차존재교다적간우신식，제출일충신영유효적건축물식별방법：수선，이용개진표기분수령산법제취건축물구역。연후，대매개분할득도적건축물구역，제취기륜곽，대륜곽진행기우평은소파변환적방사불변량계산，병구조건축물모형수거고，이용상관계수실현료건축물적유효식별。
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.