Technology Research Of Vegetation Information Extraction Based On Fully Polarimetric SAR Image

Vegetation is an important part of the ecosystem. The extracting of vegetation information is important for monitoring environmental changes to achieve sustainable development. Surface measurement and optical remote sensing technology are traditional methods to extract vegetation information. However, it is difficult to obtain accurate information of vegetation by surface measurement, and optical remote sensing is susceptible to the weather. The development of SAR technology provides a new method to extract vegetation information both timely and accurately due to its independence of weather conditions and time of the day and its penetration ability. Moreover, fully polarimetric SAR data for vegetation information extraction open a new way.This paper analyses the basic theory of polarimetric SAR, and then discusses the algorithms of extracting vegetation cover information and inversion of vegetation biomass based on the fully polarimetric SAR and validates the two algorithms. Finally, vegetation information is extracted by using the two algorithms in complex terrain area. Details are as follows:(1) Vegetation cover information is extracted by the Wishart H/α classification algorithm based on fully polarimetric SAR image. First, the fully polarimetric SAR image of Dali Lake is classified by the Wishart H/α classifier. Second, vegetation cover information is extracted by the scattering mechanism of vegetation. Third, the optical image is used to calculate the extraction accuracy of vegetation. It is found that there are large differences among the vegetation cover extraction results in different multi-look windows. The classification is improved and the accuracies of vegetation cover information extraction are increased by error analysis, with the original accuracy 54.0% improved to 73.1%, and the other accuracy 81.7% increased to 88.0%.(2) Vegetation biomass is inversed by the neural network algorithm based on MIMICS model using fully polarimetric SAR image. First, the measured growth parameters of rice in Qionglai area is inputted into the MIMICS model to generate training samples of neural network. Second, the relationship between the backscattering coefficient and biomass is established by training the neural network. Third, the well-established network is applied to a full polarimetric SAR image of Qionglai area to inverse the vegetation biomass. Finally, the accuracy of the algorithm is obtained by contrasting the inversion values with the measured inversion of values with the correlation coefficient 0.921.(3) The vegetation cover information is extracted and the vegetation biomass is inversed in complex terrain area. Maoxian mountain area is selected as the complex terrain area. DEM is used for geometric correction of the complex terrain area. The improved Wishart H/α classifier is applied to the extraction of vegetation cover, and the neural network algorithm based on MIMICS model is used to inverse the biomass of the complex terrain area to extract vegetation information in complex terrain area. Moreover, by comparing with optical remote sensing and analysing the biomass of two date fully polarimetric SAR images, the trend of biomass inversed is the same as the trend of vegetation growth. Therefore, the vegetation cover information is extracted and the biomass is inversed successfully in complex terrain area based on the fully polarimetric SAR.