On the use of ancillary data by applying the concepts of the theory of evidence to remote sensing digital image classification

This study deals with some applications of the concepts developed by the Theory of Evidence, in remote sensing digital image classification. Data from different sources are used in addition to multispectral image data in order to increase the accuracy of the thematic map. Data from different sources as well as probability images estimated from the multispectral image data are arranged in form of layers in a GIS-like structure. Layers representing belief and plausibility concerning the labeling of pixels across the image are then derived to help detect errors of omission and of commission respectively, in the thematic image. In this study, a new approach to introduce the information conveyed by belief and plausibility into the classification process is proposed and tested. Preliminary tests were performed over an area covered by natural forest of Araucaria showing some promising results.