Vegetation Spectral Feature Analysis And Water Content Inversion Based On Hyperspectral Remote Sensing Image

Hyperspectral remote sensing technology has advanced significantly in the past two decades, considerable concern has arisen over how to extract the information from the spectral of ground object or scene. However, the unprecedented spatial, temporal and spectral resolutions bring the corresponding data increase exponential increase. As data increase, so does demand for extract the effective information. This study will take vegetation as an example, commence from the spectral feature of hyperspectral, and investigate the sensitivity of reflectance to the variation in biochemical and biophysical variables of vegetation. Using the extraction of spectral feature and expression of feature parameterization, we inverted the vegetation water content and give reasonable evaluation. Finally, the method is used on the remote sensing data and gives a water content map as an application.First of all, Commencing from the vegetation spectral characteristic, this paper used the radiative transfer models PROSAIL model as a tool to simulate the vegetation spectral of different biochemistry state and investigate the sensitivity of reflectance to the variation in biochemical and biophysical variables of vegetation in 400~2500 nm wavelength range. Qualitative or quantitative analyzing explains the change trend of spectral in different biochemistry elements content. A new model named bi-inverted Gaussian model is proposed. This model can express absorption feature well and adapt the vegetation spectral. This part offer a strength theory support and technique support to the next vegetation water content inversion analyze and method.Secondly, after study and investigation of the classic vegetation water content inversion theories, including multivariate statistical analysis, methods based on spectral feature and methods based on physical model, a new method is proposed combining the vantage of classic methods. Aimed to prove the validity of the new method, an laboratory experiment is designed. The result shows new method can inverted the vegetation water content efficiently. This part will be an important support to next remote sensing application.Finally, this paper analyzes the difference between the hyperspectral remote sensing data and the spectral data obtained in the lab ground-experiments as a basement. To direct at the spectral properties of hyperspectral remote sensing data, the vegetation water content inverted method base on bi-inverted Gaussian model is improved. After segmentation of the vegetation part, the interested bands is used for analysis and water content inverted as a feasibility certification of the method and the robustness and universality. The map of vegetation water content is obtained at last, as an important application of hyperspectral remote sensing data.