A Dual-Level Hybrid Approach for Classification of Satellite Images

The traditional methods for extracting information from satellite images are generally based on the spectral response of the sensors. These approaches are in some cases insufficient in particular in case of high-resolution images. In fact these images have a spectral content increasingly heterogeneous. It is, therefore, necessary to use more efficient analysis methods. The multi-source classification is a robust analytical tool, and it is one of the most used approaches for the extraction of telemetric information. This paper is focused on the problem of the classification of satellite images by the hybridization of several methods: multi-layer perceptron, hidden Markov models and genetic algorithms. The results prove the efficiency of the proposed final approach, with a classification rate of 98.79%, significantly higher respect to the results obtained by the MLP method, and by other approaches.