A NEW MULTI-FEATURE APPROACH TO OBJECT-ORIENTED CHANGE DETECTION BASED ON FUZZY CLASSIFICATION

Remote sensing technologies have been widely used in the detection of Land Use/Land Cover change (LUCC). In the past few decades, lots of methods have been proposed attempting to detect changes using multi-temporal satellite images, most of which are on the pixel level. In this paper, a new synthetic method based on object-oriented is proposed. Several customized difference features such as difference of band value, Normalized Difference Vegetation Index (NDVI), texture and so on are applied to the change detection, and also the fuzzy classification. The classified elements are image objects with the object-oriented approach which improve the salt-and-pepper problem effectively. Experiment results show that this method has a stronger advantage than the traditional method to high resolution remote sensing image change detection.