CONCLUSIONS

The proposed HMM algorithm integrates the concept of both linear and non-linear mixing. Endmembers are first extracted from the image, eliminating the undesired spectral signatures followed by unmixing and then interpolating the fractions for the whole image. Unmixed MODIS output was compared with a HR classified output (IRS LISS-III MS) to estimate the accuracy. It showed that the overall RMSE of HMM was 0.0191 ± 0.022 as compared to the LMM output considered alone that had an overall RMSE of 0.2005 ± 0.41 indicating that individual class abundances obtained from HMM are very close to what is present on the ground and observed in the HR classified output. This emphasizes that the influence due to multiple reflections among ground cover targets has to be considered for the abundance estimation. While a linear detection method might work adequately for many scenarios, a non-linear model would perform better. The only disadvantage of this method is that the endmember selection through N-FINDR requires at least one pure pixel. Future work will involve developing methods to obtain pure pixels when there are no endmembers in the scene.