面向对象的城市绿地信息提取方法研究

Research on Detection of Urban Vegetation by Object-Oriented Classification

在比较传统的城市绿地提取方法的基础上,采用了面向对象的图像分类技术,对Quick- Bird卫星图像进行上海市区绿地信息提取实验,得到了令人满意的结果,总体分类精度达到84．4％,较传统的监督分类方法提高了24．4％,具有明显的优越性和应用前景．

The method of object-oriented classification for remote sensing images, based on image segmentation which could create objects sets of homogeneous pixels, provides a way to analyze object＇s features, such as spectral, shape, topology, texture and so on, and to realize the functions of discriminating various species and automatic classification. The traditional way of analyzing and extracting urban vegetation community was taken as a reference, a new classification method has been developed using QuickBird satellite image in Shanghai. With the new method, the total precision is 84.4 %, 24.4% higher than conventional supervised classification. The principle of the new approach mentioned may be useful as a new algorithm joined with existing classifiers.