Picea Chrenkian Var. Tianschanica Canopy Density Information Extraction Based On SAR Data

Canopy density is not only a major technical indicators of national forest resources planning and design, but also an important parameters of forest resources monitoring and evaluation. Canopy density estimation of quantitative remote sensing, in particular, the use of all-weather radar data. Have important meaning to reduce the workload of the ground survey of forest resources, improve forest resources monitoring accuracy. Taking Radarsat2 canopy density under full polarization synthetic aperture radar data quantitative estimation method as the research target, take The Picea schrenkiana var. tianschanica forest as experimental object, Radarsat fine C- 2 full polarization band image as the experimental data, To carry out the remote sensing estimation modeling of forest canopy density, inspection and evaluation results of research, So as to provide technology for forest resources investigation, monitoring method and using for reference.The paper research content and results reflected in the following three aspects:(1)Data preprocessing of The polarization synthetic aperture radar.Do multilook processing to the original SLC data in the NEST,to suppress speckle noise in images, do radiation calibration on more apparent after processing with phase and amplitude information of images, after get the scattering coefficient; and do geocoding image, to the transformate the radar coordinate system of the image to the latitude and longitude coordinates system, to overlay and forest resources data for registration, for further data mining and analysis provide information source.(2)Canopy density inversion model is established based on the backward scattering field of multi-polarization radar coefficients.On the basis of the data preprocessing, extract Backscatter coefficient of all polarization channels, Build a statistical model between the radar scattering coefficient under the multi-polar channel and the sample canopy density values, Do precision test of regression model, select the responsivity of the highest polarization modes.(3)Canopy density inversion model is established based on the polarization decomposition characteristic parameters.Use the coherent polarization decomposition model- Freeman three-component decomposition, decomposition of H/A/α. Do polarization decomposition in radar imaging research district, and obtain the corresponding polarization decomposition characteristic parameters- surface, vol, dbl, entropy, Anisotropy, alpha. On the basis of analyzing the characteristic parameters of the actual physical meaning, and to do multiple correlation analysis between each parameter, choose large component as inversion factor correlation with canopy density and canopy density linear regression model is set up, and do precision verification.Through the study:(1)based on the backward scattering coefficient under the condition of modeling, VV and HH/HV channel and the correlation of canopy density is higher, at the 0.01 level significant correlation reached more than 0.3.In all models, M5 overall accuracy is the highest, at 74.42%.(2)Based on polarization decomposition characteristic modeling conditions, Freeman- Durden surface scattering surf and H/A/α polarization decomposition of entropy H and canopy density has the highest correlation, generate the optimal model for M10, the overall accuracy of 75.33%.(3)All model, produced by the method of stepwise regression model accuracy on the whole less than backward regression model.(4)In the low level of canopy density, the optimal model for the M9, accuracy was 78.68%;Canopy density in the scale of optimal model for the M11, accuracy of 79.55%.