基于SAIL模型的光譜解混研究.pdf

1、華中科技大學(xué)碩士學(xué)位論文基于SAIL模型的光譜解混研究姓名：展昕申請學(xué)位級別：碩士專業(yè)：空間信息科學(xué)與技術(shù)指導(dǎo)教師：田巖20090527華 中 科 技 大 學(xué) 碩 士 學(xué) 位 論 文 華 中 科 技 大 學(xué) 碩 士 學(xué) 位 論 文 IIAbstract Spectra are mixed, as frequently happens in the complicated environment due to sensor res

2、olution limit. Now, with development and wide application of hyperspectral remote sensing, hyperspectral unmixing has been paid more and more attention. Not only could it compensate for the sensor resolution limit caus

3、ed by hardware, but also it could reduce costs of high resolution image acquisition. So it’s of great significance in the development of hyperspectral remote sensing. Generally speaking, there are two kinds of spectra

4、l unmixing methods: linear and nonlinear. Due to the scientificity and simplicity, linear methods have been widely used. According to the kinds of spectral mixing we proposed in this paper, linear methods are suitable f

5、or adjacent mixing. However, with respect to our research object the canopy/soil system, which is up-down mixed, linear methods are not suitable anymore. More importantly, linear methods are not based on physics but stat

6、istics. For this reason, in this paper, we use radiative transfer theory to solve spectral unmixing problem in canopy/soil system. Firstly, by the inversion of the canopy reflectance model SAIL(Scattering by Arbitrarily

7、Inclined Leaves) based on particle swarm optimizer, we obtain leaf reflectance, leaf transmittance and soil reflectance. Secondly, let the value of soil reflectance be leaf reflectance value. Lastly, we calculate pure ca

8、nopy reflectance based on SAIL. So in this way, we could obtain pure canopy spectra in the end. In view of present research working emphatically on simulated data, here we do research not only for the simulated data but

9、also for the measured data, and the experiments show spectral unmixing based on SAIL model and particle swarm optimizer is feasible. Keywords: spectral unmixing, canopy reflectance model, SAIL model, particle swarm op