Pixel-Unmixing Moderate-Resolution Remote Sensing Imagery Using Pairwise Coupling Support Vector Machines: A Case Study

A method combined with support vector machines (SVMs) and pairwise coupling (PWC) was developed to achieve land use/land cover fractions of a moderate-resolution remote sensing image. At first, SVMs were applied to solve classification problems. Then, they were extended with PWC to output probabilities as the abundance of landscape fractions. The performances were evaluated by using the “estimated” landscape class fractions from our method, fully constrained least squares method, and unmixing nonlinear SVM (u_NLSVM) method, respectively, and the results were validated by real fractions generated from the SPOT High Resolution Geometric (HRG) image. The best classification results were obtained by the proposed method, which proved the effectiveness of our method.