陆探一号全极化SAR散射特性分析与地物分类性能评估OA

Data from the LuTan-1(LT-1)full-polarization synthetic aperture radar(SAR)provide a new foundation for the land cover classification and scattering mechanism analysis by China's autonomous polarization SAR.Taking the mid-to-high latitude mixed agricultural and forestry region in Jilin and the tropical vegetation region in Hainan as the study objects,this paper presents the first systematic analysis on the scattering mechanisms,polarimetric statistical characteristics,and land cover differentiation capabilities of LT-1 full-polarization data.The experimental results demonstrate that the LT-1 data exhibit an excellent land cover discrimination ability in both typical regions.The feature selection based on the Fisher discriminant criterion reveals that:in the Jilin region,the volume scattering components and absolute backscatter intensity are dominate;in the highly fragmented environment of Hainan,the polarization ratio and normalized mechanism proportion parameters effectively mitigate the soil moisture interference,demonstrating significant discriminative value.In terms of the classification performance,the unsupervised classification in the Jilin region yields an overall accuracy(OA)of 0.785 0 and a Kappa coefficient of 0.707 1,while the random forest(RF)supervised classification improves the OA to 0.897 3 and the Kappa coefficient to 0.839 1.For the Hainan region,the unsupervised classification achieves an OA of 0.716 6 and a Kappa coefficient of 0.613 6,whereas the RF method enhances the OA to 0.867 0 and the Kappa coefficient to 0.790 5.The comprehensive analysis indicates that the LT-1 full-polarization data can effectively characterize the scattering mechanism differences of land cover across different ecological regions.The research results provide a robust technical basis for the application of LT-1 polarization SAR data in land cover classification,as well as resource and environmental monitoring.