The objective of this paper is to study several approaches to forecasting the temporal evolution of meteorological synoptic maps that carry information in visual form but without objects. Window-based descriptors are used in order to accomplish continuity so the prediction task is possible. Linear and non-linear models are applied for the prediction task, the first one being based on a spatio-temporal autoregressive (STAR) model whereas the second one is based on artificial neural networks. The method and obtained results are discussed.
展开▼