Data fusion approach for employing multiple classifiers to improve lake shoreline analysis

Abstract

Remote sensing images have been widely employed to analyze bodies of water and have become essential to studying their dynamics. While the use of indices based on the threshold segmentation technique is preferred, the search for methods that define water edge contour continues. The segmentation algorithm introduced in this study is based on Mean-Shift and Watershed methods. We propose a fusion classifier strategy which allows us to obtain results that are consistent with the segmentation process. The use of two or more segmentation processes has been shown to improve pattern recognition. It is important to implement a good data integration scheme. Preliminary results suggest that the approach reported herein can improve the definition of lake shorelines.