Forest Canopy Density Evaluation In The Mountain Areas Based On SAR Images

The development of remote sensing technology and a large number of different applications of optical, thermal infrared and microwave satellite sensors on Earth observation provided an economic, simple, efficient way for the estimation of forest canopy density on a large regional scale. For the use of optical remote sensing images in the forest canopy density estimation a variety of statistical models and physical models has been developed, but the most common models used for forest canopy density evaluation is based on the relationship between forest canopy density and vegetation spectral information. With the rapid development of synthetic aperture radar technology, which is an active microwave remote sensing and could penetrate fog and cloud to observe the earth all-day and all-weather without climate and sun’s influence in the recent year. So it received a lot of attention. Therefore, the use of different polarization characteristics of SAR images for quantitative inversion of forest parameters has important application prospect.This thesis focuses on forest canopy density evalution in the mountain areas based on SAR images. Envisat ASAR image and Google high-resolution images are used in the research, and some ideas are provided to invert the forest parameters based on synthetic aperture radar technology. Some aspects are included in this thesis as follows.Firstly, Envisat ASAR satellite image preprocessing is carrying out in Mao Xian Longmen Shan area, because of the great topographic relief in the southwestern China mountains. So a mask processing is made in the preocessing. Then Google high resolution optical images is combined with non-layover, non-shaded SAR image to extract forest canopy density samples.Secondly, forest canopy density samples are extracted from Google high resolution optical image as observations, then relationships equation is modelled between backscattering coefficient of SAR image and forest canopy observations. Taking into account of the study range in a mountainous area with great topographic relief, thus, the backscatter characteristics would be affected by terrain and radar incidence angle. The stepwise regression analysis is used to investiagte these factors and the backscattering coefficient factors as independent variables. We build a statistical model with canopy density and use cross-validation method to evaluate the accuracy of the models to obtain the best statistical model.Thirdly, on the basis of KARAM scattering model, using C#language, we developped the software system to simulate direct scattering value of forest canopy and inverse forest canopy density based on the backscattering coefficient of image, and analyzed their relationship. As follows, we obtained a few main conclusions by studying:Firstly, there are good linear or non-linear relationships between canopy density and multiple radar signatures when canopy density is greater than0.4. And the correlation between canopy density and HH polarization or HV/HH polarization ratio is better than HV polarization.Secondly, The terrain and radar incidence angle had no significant effect on the relationship between HH polarization and canopy density. However, the radar incidence angle has effect on the relationship between HV polarization and HV/HH polarization ratio. The HH polarization and HV/HH polarization ratio can be treated as the independent variable parameters. We obtain a linear and exponential regression equations with canopy density and use leave-one-out cross validation to evaluate the accuracy of the established model. The results show that the model accuracy of HH polarization index regression equation is highest (R=0.79627, RMSEP=0.03602). When the canopy density is less than0.4, the correlation is poor between the image backscattering coefficients and canopy density.Thirdly, here is a strong correlation between cover and forest canopy direct backscattering coefficient value which simulated by the KARAM scattering model system. The fitting accuracy of its optimal exponential regression equation is the best, the coefficient of determination R2=0.74706. This thesis obtained the canopy direct backscatter by building statistical model between canopy density and other factors scattering values, then inverse the canopy density of forest samples. The results show that, the canopy density estimation values calculated by the method is close to the observation canopy density (RMSE=0.0561), and there is a strong correlation between them (R=0.7155)...