Remote sensing image classification by Bayesian network classifier based on causality

It has always been a hotspot and difficult point in remote sensing to identify interesting geographical objects from remote sensing images. To reduce the independence between the random variables in the network Bayesian classifier model and to improve the classification performance, a causality-based network Bayesian classifier is suggested in this paper. In this model, the improved genetic algorithm is used for network topology learning and causality analyzing is taken for feature selection, which aims to realizing automatic recognition of unfavorable geological bodies. Experiments show that this model is of good classification performance and of high classification stability. © 2014. IFSA Publishing, S. L.