Método multi-resolução adaptativo para classificação simultânea de áreas rurais e urbanas

The present work seeks to develop an adaptive classification method operating simultaneously with different rules in images including rural and urban areas. Traditional techniques are commonly based on the application of only one type of classification strategy. This assumption generally causes bad fitting of some areas if the image includes different kinds of targets. The aim is to obtain a more efficient thematic classification through the combined using of techniques and data of different resolutions, when compared with results achieved using a single approach. The proposed formulation is based on the premise that classification of rural and urban targets usually show large variations depending on how they are classified. Thus, traditional classifiers applied in environments that include rural and urban areas eventually end up benefiting one area over another. The first step of the suggested technique performs a prior automatic separation of urban and rural targets from the studied area, which will then be classified with different methods and input data. One experiment was performed using data with compatible resolution and classification techniques, according to the literature. Visual comparisons with classifications made only by means of one type of classification strategy leads us to visually verify the soundness of the suggested framework.