THE NEW APPROACH TO APPLYING THE DEZERT-SMARANDACHE THEORY IN LAND-COVER CLASSIFICATION IN UAV-BASED REMOTE SENSING

Abstract

Nowadays UAV-based Remote Sensing gives a new opportunities for conducting scientific research in a much more detail way. Classification is one of the most important procedures in Remote Sensing tasks. This procedure can be applied in solution of numerous ecological and practical tasks, such as: forest classification, determing of soil types, exploring of oil and gas. Classification of incomplete, imprecise and high conflicting data has always been and still remains the one of most important procedures of remote sensing. In this paper the new approach to applying Dezert-Smarandache Theory in UAV-based remote sensing problems is proposed. This theory can deal with imprecise and highly conflicting information. Dezert-Smarandache Theory and its main concepts are compared with Dempster-Shafer Theory and Probability Theory. Main advantages and disadvantages of these theories were described and analyzed in this paper. The development of Dezert-Smarandache Theory arises from the necessity to overcome the limitations of Dempster-Shafer Theory and Probability Theory. Dempster-Shafer Theory deals only with exhaustive and exclusive hypotheses, that can sometimes provide wrong results of classification. But exhaustive hypotheses can potentially overlap, they can not properly identified and defined. Dezert-Smarandache Theory deals with exhaustive and not mutually exclusive hypotheses. This theory deals with all hypotheses, all intersections and unions of these hypotheses. It is shown, that Dezert-Smarandache Theory can solve classification tasks more efficiently than Probability Theory and Dempster-Shafer Theory, when we combine imprecise and high conflicting data. It is described classical Dezert-Smarandache  fusion rule in this paper. It is also considered numerical example with applying Dezert-Smarandache fusion rule for classification of multispectral aerospace images. The proposed approach to applying the Dezert-Smarandache Theory in land-cover classification can be applied in different agriculture and practical tasks, ecological monitoring and searching for minerals.