Typical approach to classification of objects on remote sensing images, which assume detection of terrain details being grouped into several thematic categories, represented by radiometric properties of the multispectral raster image, still needs more and more sophisticated methods of increasing efficiency. Among currently used solutions, on special attention claimed methods stated as the object-oriented, which are based on analyses of parts of the bitmap, grouped according to criteria of homogeneity. It is possible to carry out the image segmentation using various methods, which are developed within couple of informatics’ sciences. In interpretation of remote sensing images there were elaborated approaches adapted to specific of those types of images. Such special use of grouping technique of objects fulfilled the particular accuracy criteria were investigated here. There were described the density-based clustering method of large databases and their implementation adapted to raster image analyses, and then, the possible solution using accessible vector information was suggested. The text is illustrated using the simplified model of remote sensing image.

29-39