Evidence Theory And Its Application In Image Recognition

Evidence reasoning has perfect performance in expression of uncertain knowledge, which is the reason why it has been making great progress in theory and application in recent years. It is, however, an obstacle how to combine the evidences which conflict with each other. To overcome it, some theory and application works are done in this paper.Firstly, the latest theory and application advances are summarized. The essence of each concept is analyzed carefully, and related formulae linking these concepts are proven. New rules to deal with conflicting and correlative evidences are given. Some approaches to decrease computation burden and to combine fuzzy and priori evidences are summed up.Secondly, absorptive method and weighed distribution have been expanded and their validities are proven. It is proven that the belief degree of the main focal element increases after combining two conformable evidences and decreases after combining two conflicting evidences. Quantitative results of existed combination rules' robust region are deduced according to conclusion above.Thirdly, evidence reasoning is applied to image recognition. A method of improving belief degree by use of evidence reasoning to combine multiple images arising from the same target is studied. Two methods of forming basic belief assignment function are proposed, and recognition effect is compared by using various combination rules and by using various moments with the same combination rules. The result indicates that evidence reasoning seems to be another effective approach to target recognition.Finally, in order to make further research, some prospects of evidence theory are proposed.