摘要
本文利用高光谱遥感影像和机载LiDAR数据融合来重点提取城市的房屋和树木目标。首先通过机载LiDAR点云数据提取数字地面模型(Digital Surface Model,DSM)和数字地形模型(Digital Terrain Model,DTM),进一步差值计算得到nDSM(normalized Digital Surface Model)。利用高光谱影像计算归一化植被指数(Normalized Difference Vegetation Index,NDVI),采用主成分分析法(Principal Component Analysis,PCA)进行去噪和降维处理。结合nDSM和NDVI图像,采用面向对象的特征提取方法实现研究区内大型商用房屋的提取。对PCA图像和nDSM图像进行融合,然后采用最大似然分类(Maximum Likelihood Classification,MLC)方法进行监督分类,实现民用房屋和树木的提取。本文研究结果显示:商用房屋的正确提取率达到89.53%;MLC方法对融合图像分类的总体精度为84.00%,Kappa系数为82.86。
High spatial resolution hyperspectral image and airborne LiDAR data are very useful in urban applications.In this paper,the fusion of hyperspectral image and airborne LiDAR data was implemented to extract buildings and trees in urban area. Digital Surface Model (DSM)and Digital Terrain Model (DTM)were derived from airborne LiDAR point cloud data,and then normalized Digital Surface Model (nDSM)was computed from the difference of DSM and DTM.Normalized Difference Vegetation Index (NDVI)was computed from hyperspectral image using near infrared (NIR)and red (R)bands.Principal Components Analysis (PCA)transformation was used to reduce the noise and dimension of the hyperspectral image.Rule-based object-oriented feature extraction method was used to extract commercial buildings from the combination of nDSM and NDVI data.Supervised classification based on Maximum Likelihood Classification (MLC)method was used to the fused image of PCA and nDSM for the extraction of residential buildings and trees.Results showed that commercial buildings were extracted with high accuracy of 89.53%.MLC was the more effective classifier to the fused image,with overall accuracy of 84.00% and Kappa coefficient of 82.86,respectively.
出处
《遥感信息》
CSCD
2014年第6期73-76,83,共5页
Remote Sensing Information
基金
中国地震局地震预测研究所基本科研业务费专项(2010IES0203)
国家863计划项目(2012AA12A306)