基于小波变换的卫星遥感地图中建筑物识别  

Research on Building Recognition from Satellite Remote Sensing Image Based on Wavelet

作　　者：赵洁[1] 李光耀[1] 庞池海[1] 朱恒晔[1] 

机构地区：[1]同济大学CAD研究中心,上海201804

出　　处：《计算机技术与发展》2008年第11期243-246,共4页Computer Technology and Development

摘　　要：建筑物作为城市的一个重要特征,对它的识别已经被广泛应用在城市仿真等许多领域。目前,传统方法通常采用基于建筑物边缘线性特征的识别算法,虽然该类方法具有简单高效的特点,但识别率较低。提出了一套基于小波变换的建筑物识别方法。首先,利用二进小波变换检测建筑物边缘,然后通过计算基于小波的仿射不变量识别建筑物,最后构建了建筑物轮廓模式数据库。实验结果表明所提方法能够准确有效地识别卫星遥感地图中的建筑物。Building is an essential characteristic of a city. The recognition of it has been applied broadly in many fields like urban simulation. Utilizing linear feature of building border was now widely used in traditional building recognition. Though simple, the recognition accuracy of these methods was low. A novel method of recognizing building from remote sensing images based on wavelet analysis and affine invariant was proposed. Firstly, dyadic wavelet was used to extract the edge, and then buildings were recognized by computing the affine invariants, In addition, a building recognition pattern database was constructed. Experimental results show that the proposed method can recognize building from remote sensing images both correctly and efficiently.

关 键 词：建筑物识别 卫星遥感地图 小波变换 仿射不变量 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]