基于多源遥感数据的城市不透水层估算方法研究

项目名称： 基于多源遥感数据的城市不透水层估算方法研究

项目编号： No.41201357

项目类型： 青年科学基金项目

立项/批准年度： 2013

项目学科： 地理学

项目作者： 孙中昶

作者单位： 中国科学院遥感与数字地球研究所

项目金额： 25万元

中文摘要： 城市不透水层不仅是城市化进程的指示器，更是评估城市环境的一个重要敏感因子，因此开展不透水层及其相关方面的研究已成为当前研究的热点。传统的不透水层提取大多基于中、高分辨率光学遥感影像，但存在很多限制，主要包括：异物同谱、混合像元、阴影以及树冠覆盖下的不透水层探测等。为解决这些问题，本项目采用全极化数据、机载激光雷达数据以及高光谱影像等多源遥感数据相结合来发展精确的城市不透水层估算方法。本项目首先基于极化数据进行不透水层估算方法研究，重点开展极化数据不透水层估算模型、极化特征优化选择、利用升降轨消除阴影和叠掩；进一步融合光学影像提高不透水层估算精度；然后探讨综合高光谱和LiDAR的高精度不透水层估算方法；并且将PSVM算法应用到不透水层估算中；最后利用高分影像和实测数据进行不透水层精度评估。研究结果为高精度不透水层遥感估算提供一些新方法，进而推动极化雷达和LiDAR在不透水层遥感估算中的应用。

中文关键词： 极化合成孔径雷达；激光雷达；城市不透水层；支持向量机；监督分类

英文摘要： Urban impervious surface is not only an indicator of the degree of urbanization, but also a major sensitive factor of urban environment.The impervious surface is closely related to many environmental problems, such as water quality, stream health, the urban heat island effect and so on. Therefore, impervious surface and its effects on urban environment have attracted more interest recently in the remote sensing community.Rencently, urban impervious surface is mostly extracted from medium-high resolusion optical remote sensing imagery, and many information extraction methods are developed based on optical imagery. However,there are some limitations for estimating impervious surface from optical imagery, including different objects with same spectrum, mixed-pixel, shadows caused by tall buildings or large tree crowns, and extrating impervious surfaces covered by tree crowns.In order to resolve those limitations, multi-source remote sensing data (e.g. PolSAR data, airborne LiDAR data and optical imagery) are used to develop the estimation methods of high-precision urban impervious suface in our project. Firstly，this project investigates the impervious surface estimation method using PolSAR data, and focuses on researching impervious suface estimation model based upon PolSAR data,PolSAR features optimizing selection

英文关键词： Polarimetric synthetic aperture radar (PolSAR)；Lidar；urban impervious surface；support vector machine (SVM)；supervised classification