期刊名称：ISPRS Annals of the Photogrammetry, Remote Sensing and Spatial Information Sciences
印刷版ISSN：2194-9042
电子版ISSN：2194-9050
出版年度：2010
卷号：XXXVIII - 4/C7
出版社：Copernicus Publications
摘要：The Pantanal (Brazil) is one of the world's largest tropical wetlands. A natural phenomenon that encourages more detailed scientific studies on this region is the genesis and behavior of innumerable small lakes and ponds with open-water surfaces on the Nhecolandia region at the Taquarí alluvial fan. In this paper, we suggest a new approach to detect and differentiate lakes among saline and non- saline using solely remote sensing data and object-based image analysis, which have still been undertaken towards this end. In all experiments the visible and near-infrared bands of the ASTER sensor were used. The methodology involved segmentation steps using the multi-resolution segmentation algorithm available on the Definiens Developer system as well as rule-based classifications and merging of objects. The classification model was structured as s process-tree. Initially, a segmentation and classification step was carried out with the purpose of precise detection and delineation of all lakes. Following, sand belts were detected also through a segmentation and classification process. Finally, the differentiation between saline and non-saline lakes was done based on the criterion of relative border to sand belts. A qualitative analysis of the resulting thematic map indicated a good result. Quantitatively, the error matrix scored an overall accuracy of 73% as well as a Kappa index of 0.64. Most of the classification errors were ascribed to intermittent lakes on different drought conditions. The object-based classification approach has proven to be very efficient for the detection of lakes and their differentiation between saline and non-saline in the region of Pantanal of Nhecolandia
关键词：Pantanal of Nhecolandia; Object-based Image Analysis; Lakes Delineation; Rule-based Classification
