Crop Classification Based On SAR Image With Polarization Characteristics

Polarimetric Synthetic Aperture Radar(Pol SAR)can characterize the structure and dielectric properties of the target surface,and is one of the important methods for classification of ground objects.The reality of the foundation.For agricultural and forestry crops that exhibit different scattering characteristics at different growth stages,single-phase dual-polarization SAR images often cannot provide sufficient growth information,so multi-phase dual-polarization SAR data is more conducive to the classification of agricultural and forestry crops.This paper proposes a classification method of agricultural and forestry crops based on polarized growth characteristics.The backscattering coefficient and H/α polarization decomposition parameters are weighted and merged.The obtained polarized growth characteristics can accurately represent the growth law of the target.The classification is implemented in the recurrent neural network to verify the ability of polarized growth characteristics in crop classification.The main work and innovation are as follows:(1)Comprehensive analysis of time-varying characteristic data of agricultural and forestry crops.Aiming at the problem that the backscattering coefficient characterizes the crop biomass without considering the main scattering mechanism changes,the H/α decomposition feature has the advantage of regular changes on the H/α classification plane,and the backscattering coefficient and H/αdecomposition parameters are the characteristics.Establish the polarization scattering index time series model to realize the fusion of two types of features.The results show that the polarization scattering index can reflect the changing trend of ground objects in different time phases.(2)Multi-temporal dual-polarization SAR agricultural and forestry crop classification method based on polarization growth characteristics.In view of the situation that the scattering value of some ground objects will oscillate sharply in some time phases,etc.,this article first performs data integration on the backscattering coefficient and H/α decomposition characteristics under two polarization modes,and then uses the ideal growth curve The model fuses these features and weights all the features through the goal planning method,so that the fused feature trend can more accurately represent the growth characteristics of agricultural and forestry crops.Experimental results show that compared with backward scattering coefficient,polarization decomposition parameters,their joint and the polarization scattering index,this paper proposed polarization feature fusion classification accuracy is higher,growth of rice,forests and bare the overall accuracy of 90%,rice classification accuracy reached 95%,and show that this method is able to blend into the advantage of the scattering coefficient and the decomposition of H/alpha parameters,,and better describe the change process of the growth of agricultural and forestry crops.