Surface Parameters Inversion Based On Fully Bistatic Radar Scattering Feature Map

High-precision and large-scale surface observation data,including surface temperature,soil moisture,snow,vegetation,surface roughness,etc.is required for research and applications in many fields such as hydrology,meteorology and agriculture require.Among them,soil moisture and surface roughness,as important surface parameters reflecting the surface conditions,have important applications for hydrology and water resources,meteorology and climate,and agricultural research.Therefore,the inversion of highprecision surface parameter is of great significance.Microwave remote sensing technology has become an important monitoring method for studying surface characteristics due to its sensitivity to a variety of surface element characteristics.And with the rapid development of microwave remote sensing,it has been widely used in monitoring and quantitative inversion of surface elements.Bistatic scattering can provide more scattering information data.Radar data,with its unique polarization,sensitivity to ground surface parameters,and phase characteristics,have been considered to be one of the most effective methods for monitoring surface parameters with high spatial resolution.In this paper,the theoretical scattering model AIEM model and neural network are combined to carry out the inversion of surface parameters.The research is divided into two parts: one is the bistatic polarization scattering hemisphere diagram,and two-dimensional curve diagram of the relationship between parameters and scattering coefficients,we analyze the sensitivity of radar and surface parameters that affect the surface scattering characteristics,and find linear or non-linear relationship between radar scattering coefficients and the surface parameters;the other is to use two structures(traditional structure and residual structure)of convolutional neural network to invert surface parameters,using data of AIEM simulation as sample input,the framework of ground parameter inversion model based on convolutional neural network is constructed.Since the accuracy of the convolutional neural network mainly depends on the training samples and test samples,the database is brought into the convolutional neural network model of two structures.After repeated training,a highprecision surface parameter inversion model is obtained.For traditional neural networks to compare backward and bistatic,the inversion accuracy of bistatic scattering data is obviously higher than that of backscattering data.After 6750 trainings,the accuracy of dual station fully polarized scattering data is reaches a highest level.The same results can be obtained for residual convolutional neural networks.By comparing the accuracy of the three sets of data(back co-polarization,bistatic co-polarization,bistatic full polarization),the following conclusions can be drawn: the accuracy of the bistatic scattering coefficient simulated by the AIEM model is higher than the backscattering coefficient,and the inversion accuracy of by bistatic full-polarization data is higher than that of bistatic homo-polarization data.Therefore,bistatic radar has better performance in retrieving surface parameters than mono-station radar.