证据理论结合遥感分类数据能力定量评价研究

DS(Dempster-Shafer)证据理论具有结合多源数据的能力,在遥感分类中应用越来越广泛.然而,并不是所有数据源利用证据理论结合后都能提高目标类别的基本概率分配(Basic Probability Assignment,BPA),从而提高遥感分类效果.如何对证据结合的效果进行评价已成为应用证据理论的一个关键问题...

The Quantitative Evaluation of Remote Sensing Data for Supervised Evidential Classification

 DS (Dempster-Shafer) evidence theory has the capability of combining multisource data，and has been used more and more widely in the remote sensing classification field. However，it is not true that all the data sources can improve target category’s Basic Probability Assignment (BPA) so as to improve the remote sensing classification accuracy. The evaluation of the effect of combined evidence is therefore a key point in the application of DS evidence theory. This paper proposed the evidence combine index (eci) for evaluating the combined evidence. The authors chose band 5 and band 7 of TM image as verification data，applied the eci index to evaluate the combining effect，and used the variation of kappa coefficient before and after evidence combination classification to validate the eci. The results show that the eci index can reflect the effect of evidence combination and thus lay the foundation for evaluating supervised evidential classification quantitatively.