基于极化SAR和光学影像特征的土地覆盖分类

Land cover classification based on polarization SAR and optical image features

为了更好地利用多源遥感影像参与土地覆盖分类,采用一种基于合成孔径雷达(SAR)影像和光学影像相结合的特征分析及筛选方法.该方法在对光学影像和SAR影像的各类特征变量进行可分性分析后,使用随机森林算法对其组成的高维变量空间进行降维筛选,将筛选出的特征变量用于土地覆盖分类,并对实验结果进行分析比较.实验结果表明:利用随机森林算法对特征变量进行分析筛选后的变量组合可以取得最优的分类结果,总体精度和Kappa系数可以达到92.1%和0.91,相比于仅用SAR影像特征变量进行分类时分别提升了11.9%和16.7%.该方法能够充分发挥光学影像和SAR影像各自的优势,提高特征变量的利用率,使分类结果更加稳定和精确.

In order to make better use of multi-source remote sensing images to participate in land cover classification,a feature analysis and screening method based on the combination of synthetic aperture radar(SAR)images and optical images is adopted.After analyzing the separability of various feature variables of the optical image and SAR image,the random forest algorithm is used to reduce the dimensionality of its high-dimensional variable space.The selected feature variables are used in land cover classification,and the experimental results are analyzed and compared.Experimental results show that the variable combination after the analysis and screening of the characteristic variables using the random forest algorithm can obtain the best classification results.The overall accuracy and Kappa coefficient can reach 92.1%and 0.91.Compared with the classification using only the SAR image characteristic variables,they increase by 11.9%and 16.7%,respectively.The proposed method can give full play to the respective advantages of optical and SAR images,improves the utilization of feature variables,and makes the classification results more stable and accurate.