Decision Fusion Approaches to Spatiotemporal Change Prediction in Satellite Images: A Comparative Study
Résumé
Predicting changes in satellite images remains one of the major challenges in the remote sensing field. However, this process is usually accompanied by several imperfections leading sometimes to erroneous decisions. In order to solve this problem, we propose to introduce knowledge discovery methods to improve the prediction process in satellite images and to decrease the associated imperfections. The proposed approach addresses this issue through the use of two fusion methods which are the evidence and adaptive possibility methods. Performances of these two methods were evaluated using several time series of SPOT-5 images. The obtained results show that the adaptive possibility method outperforms the evidence method in predicting land cover changes in most cases.