Wetland classification using evidential reasoning approach with multi-temporal landsat ETM+imagery

The Zhalong National Nature Reserve (NNR) that located in Heilongjiang Province was selected as a study area, five temporal ETM+imagery covering the study area were used as material for wetland mapping. An evidential reasoning (ER) approach that built on Dempster-Shafer theory was introduced to combine the five images. We used a new algorithm named minimum distance to generate evidence support from ETM+image. Results indicated that ER was an effective approach to integrate multi-temporal imagery for wetland mapping. Compared to classification results of single image based maximum likelihood classifier (MLC), the improvement of overall accuracy by ER approach was apparent (2%-12%).