APLICAÇÃO DE TÉCNICAS DE FUSÃO DE IMAGENS, SEGMENTAÇÃO E CLASSIFICAÇÃO POR REGIÕES COMO UMA NOVA ABORDAGEM PARA O MAPEAMENTO TEMÁTICO DETALHADO DE ASSOCIAÇÕES SOLO-VEGETAÇÃO

The paper discusses the combined use of image merge, segmentation and region-classification techniques, as a new approach in the semi-automatic mapping of land-cover types. In the first step of the procedure, a digitized panchromatic aerial photograph was co-registered with Landsat-TM images. A hybrid image set with high-spatial resolution was then produced by merging the Landsat images and the aerial photograph, through intensity (I), hue (H) and saturation (S) color transform. Using segmentation techniques, hybrid images were partitioned off into homogeneous regions, and classified according to a region-based classification algorithm. The analysis of the hybrid IHS color composite supported by field data information permitted to identify the classes on the classified image, so producing an accurate thematic mapping of different soil-vegetation assemblages in the study area. Enhanced precision and time saving are the main advantages of this approach, in comparison with the conventional visual interpretation.