Apply semi-supervised support vector regression for remote sensing water quality retrieving

This paper proposes a novel semi-supervised regression model with co-training algorithm based on support vector machines, which retrieves water quality variables from SPOT5 remote sensing data. Nonlinear relationship between water quality variables and SPOT5 spectrum are described by two support vector regression (SVR) models. Semi-supervised co-training algorithm for the two SVR models is established. The method is used for retrieving four representative water quality variables of the Weihe River in Shaanxi Province, China. The results show that the new method has better performance than the statistical regression method. Through integrating two SVR models and using unlabeled samples, an operational method when paired samples are limited is obtained. Combining techniques of machine learning and remote sensing, it provides an effective approach for remote sensing water quality retrieving.