融合WorldView-2与机载激光雷达数据估算城市不透水面

Synergistic Use of WorldView-2 Imagery and Airborne LiDAR Data for Urban Impervious Surface Estimation

针对高空间分辨率遥感影像进行城市不透水面提取时存在的同物异谱、异物同谱及阴影等局限性,提出一种基于WorldView-2高分影像与机载激光雷达数据融合的分层分类估算城市不透水面的方法。该方法首先运用基于雾霾与比值(haze-and-ratio-based,HR)的融合算法对WorldView-2多光谱波段与全色波段进行数据融合;然后依据LiDAR归一化数字表面模型(normalization digital surface model,nDSM)高度阈值分为地面物体与非地面物体,运用像元尺度上分层支持向量机分类算法进行城市不透水面百分比估算;最后结合特征阈值和GIS空间分析法探测阴影区域不透水面。研究结果表明,与传统的高分影像提取城市不透水面方法相比,该方法可以明显改善材质复杂的建筑物屋顶提取不完整,以及高亮裸土与高反照度屋顶相互混淆的现象,并通过阴影校正可以较好地区分阴影区域的植被与不透水面信息,进而提高城市不透水面估算精度。

There are lots of challenges for deriving urban impervious surface for high spatial resolution imagery,including spectral similarity of different objects,same objects having different spectral characters,and shadows of tall buildings and large tree crowns.In order to reduce these uncertainties,a hierarchical classification and urban impervious surface estimation method was proposed in this study,using WorldView-2high-resolution imagery and LiDAR data.In the study,firstly,a Haze-and-Ratiobased（HR）fused scheme was used by WorldView-2 Multispectral（MS）and Panchromatic（PAN）images.Secondly,ground objects and non-ground objects were split to two layers based on LiDAR normalization Digital Surface Model（nDSM）threshold.Thirdly,a hierarchical and pixel-based Support Vector Machine（SVM）classification was applied to estimate impervious surface in ground objects and non-ground objects.Finally,a shadowed impervious surface detection method was proposed,combining spectral and height threshold with GIS spatial analysis.The results showed that,compared to the traditional method from high resolution imagery,the proposed method improved the impervious surface extraction,especially complex rooftops,and reduced a confusion between dry soil and high albedo rooftops.Shadow correction could reduce the overestimation of impervious surface percentage and improve the accuracy of impervious surface estimation.