Comparison and improvement of Dempster-Shafer models

In this paper, a performance comparison of different classification algorithms based on the evidence theory of Dempster-Shafer will be presented. The comparison is illustrated on Landsat Multispectral Scanner (Landsat MSS) data whose ground truth is provided. A modification of the Denoeux's model [1] is proposed showing better and more stable classification results of the Landsat MSS data.