Evaluation of different change detection methods in the semi humid tropics, possibilities and limitations

摘要

Detection of land use / land cover changes is an important task for remote sensing. To compare different change detection methods, principal component analyses (PCA) and change vector analyses (CVA) were performed on multi-temporal LANDSAT data. The test site is situated within the semi humid tropics in central Benin (West-Africa). In this paper important steps like pre-processing of the images and an objective method for derivation of thresholds are described. The results of the different change detection methods are discussed intensively and compared with each other. It turned out that the PCA is more stable against atmospheric influence and performs well in detecting major changes , but the CVA is more easy to interpret and performs better detecting slight changes, as well it gives information about the direction of the changes. As an important outcome of this work it must be stated that the pre-processing of the data, especially the atmospheric correction has major influence on the results of the change detection.