基于LiDAR和CCD数据的地形与建筑提取方法优化及精度评价

借助机载小光斑激光雷达点云数据,采用Kraus滤波法结合增强Canny算子优化提取数字高程模型,然后结合LiDAR数据中提取的nDSM和粗糙度特征,以及CCD数据中获得的光谱属性和几何属性,应用多源特征融合面向对象影像分类方法,以提高城市环境下遥感分类的可靠性和建筑实体信息提取精度。结果表明:DEM估测值变异解释能力达到96%,其均方根误差1.15 m,拟合的直线紧贴1∶1线;同时,结合粗糙度、光谱信息和形态指数等信息分类的方法不仅缓解了分类“噪声”,降低了错分现象,且精度较高;研究区内建筑的面积决定系数大多高于0.7,高度信息的估测值变异解释能力也均达到92%以上,表明基于多源特征融合的面向对象分类方法结果可靠且对建筑的三维结构参数提取精度高。

Method Optimization and Accuracy Evaluation of Terrain and Buildings Extraction based on LiDAR and CCD Data

In this paper,Kraus filtering algorithm and enhanced Canny arithmetic operators were applied to extract digital elevation models by small foot\|print LiDAR point clouds;in order to enhance the urban remote sensing classification reliability and ground features extraction accuracy,the buildings and other features were extracted by objected-orientated classification method with merging multi\|sources of spatial data,through the integration of LiDAR-extracted nDSM & roughness index and CCD-derived spectral & geometrical attributes.The result demonstrated that:the estimated DEM accounted for around 96% of the variation with a standard error of 1.15m,and the two estimates were not significantly different from a 1∶1 relationship;meanwhile,the information classification method,which integrates roughness index,spectral information and shape index,could not only relief the noise of classification and reduce the classification error,but also reach a relatively high accuracy.The coefficiency of  determination of the buildings areas was almost higher than 0.7,and the explanation capability of the estimated buildings heights accounted for above 92% of the variation,which showed that the result of the objected\|orientated classification by merging multi-sources was reliable for extracting 3D structure parameters of the building with a high accuracy.