Abstract—Land cover (LC) refers to the physical state of the Earth's surface in terms of natural environment such as soil, vegetation, water, etc. Since most LC features occur at spatial scales much finer than the resolution of the primary satellites, LC mapping at subpixel level is required to obtain abundance maps of each category in a given pixel. These abundance or fractional maps are obtained using linear mixture model, which assumes no interaction between materials and a pixel is treated as a linear combination of signatures with relative concentrations. The model allows a number of different LC types to be present, each contributing a fraction of its unique spectrum corresponding to the area occupied by that LC type in a pixel. The linear model is inverted to produce estimates of those fractional abundances. The optimal solution of the mixture models can be either an unconstrained or partially constrained or a fully constrained solution depending on whether the constraints have been imposed. In this paper, we discuss the implementation and comparative analysis of an unconstrained, partially constrained and fully constrained least squares linear mixture models for LC class quantification using a series of computer simulations and Landsat data experiments to assess their performance.

 

Index Terms—land cover, linear mixture model, endmember, subpixel classification, algorithm

Cite: Uttam Kumar, Cristina Milesi, Ramakrishna R. Nemani, S. Kumar Raja, Weile Wang, and Sangram Ganguly, "Partially and Fully Constrained Least Squares Linear Spectral Mixture Models for Subpixel Land Cover Classification Using Landsat Data," International Journal of Signal Processing Systems, Vol. 4, No. 3, pp. 245-251, June 2016. doi: 10.18178/ijsps.4.3.245-251