A Multiple-Band Algorithm for Separating Land Surface Emissivity and Temperature from ASTER Imagery

We intend to propose a multiple-band algorithm which can simultaneously retrieve land surface temperature and emissivity from ASTER data. We build four radiance transfer equations for ASTER band 11, 12, 13, 14, which involve six unknown parameters (average atmosphere temperature, land surface temperature and four bands emissivity). We also analyze the emissivity characteristics of common objects about 160 kinds provided by JPL spectral database between thermal band 11, 12, 13, 14 and find that there is approximate linear relationship between them. For common 80 kinds terrors, the average emissivities error of band 11 and 14 are all under 0.01, the max emissivity error is under 0.0097 for band 11 and 14. So we can obtain six equations and six unknown parameters. In order to improve the accuracy, we can make some classification before retrieving land surface temperature. We can use three methods to resolve the equations. The first is that we make classification for image and get different equation, then resolve the equation. The second is Least-squares. The third is that, we can simulate database according to the characteristics of objects and utilize the neural network to resolve equations. The analysis indicates that the neural network can improve the practical and accuracy of algorithm.