特征筛选方法在高分辨率遥感影像建筑物提取中的应用

Application of Feature Selection Method in Building Information Extracting from High Resolution Remote Sensing Image

城市建筑物信息的自动提取是城市遥感的关键技术之一,由于阴影、下垫面等多因素干扰,建筑物信息提取精度往往不稳定。本文以Pleiades卫星影像为数据源,通过改进Relief F特征筛选方法,探讨建筑物信息提取精度提高的可行性。首先构建高分辨率遥感影像建筑物基础特征空间,然后利用改进型Relief F算法分析特征对象的权重并筛选出最优特征,最后用监督分类、无特征筛选分类和基于改进型Relief F特征筛选等3种方法分别提取研究区建筑物信息,并结合实地调查数据进行精度验证。结果表明,基于改进型Relief F特征筛选的分类方法提取精度能够达到91.34%,较其他两种方法提取精度分别提高了34.31%和5.62%,且运算速度快,自动识别效率高。

The automatic extraction of urban building information is one of the key technologies of urban remote sensing,however,the accuracy of extraction is often unstable due to many factors such as shadows and undersides. Taking the Pleiades satellite image as a basic data source,the paper explores the feasibility of urban building information extraction by advanced Relief F feature selection method.Firstly,the basic feature space of buildings with high resolution remote sensing image is built. Then,the optimal characteristics are selected by the weights which are determined by advanced Relief F algorithm. Finally,the supervised classification,the non-feature selection classification,and the advanced Relief F feature selection methods are respectively used to extract the building information in the study area,and the accuracy of the extraction is verified and compared by site investigation data.The result shows that the advanced Relief F method can reach a higher accuracy with 91.34% and a higher extraction speed with 34.31% and 5.62% higher than the other two methods.The result demonstrates that the advanced Relief F feature selection method has certain reliability and applicability,and can make the work of building information extraction become more automatic and intelligent.