融合机载LiDAR和高光谱影像的土地利用分类

Fusion of airborne LiDAR and hyperspectral image for land use and cover classification

高光谱影像具有丰富的波谱和纹理信息,机载LiDAR点云数据包含了地物高密度、高精度的三维信息。分别从两种数据中提取地物的光谱特征、纹理特征和高度特征,并进行不同的特征组合,然后采用随机森林分类器进行地物分类实验。结果表明,机载LiDAR点云和高光谱数据在地物分类方面具有很强的信息互补性;融合了LiDAR高度特征的总体分类精度和Kappa系数均优于仅使用高光谱影像,其中“PCA+NDVI+GLCM+CHM”的特征组合总体分类精度和Kappa系数最高,分别为85.96%和0.81;与未加入LiDAR特征的组合相比,总体分类精度提高了5.33%。

Hyperspectral images contain abundant spectral texture information,airborne LiDAR contain high-density and high-precision three-dimensional information of the features.In this study,based on airborne LiDAR data and hyperspectral data,height features,spectral features and texture features are extracted from the two kinds of data respectively to form different feature combinations,and then random forest classifier is used for land use and cover classification.The results indicate that airborne LiDAR and hyperspectral data have complementary information in terms of feature classification;the overall classification accuracy and Kappa coefficient of the combination of LiDAR height features were better than those of hyperspectral images only,with the highest overall classification accuracy and Kappa coefficient of 85.96%and 0.81 for the combination of"PCA+NDVI+GLCM+CHM",respectively;Compared with the combination without the LiDAR features,the overall classification accuracy was improved by 5.33%.