"FUSION FOR CLASSES IN DIFFICULTY" FOR ACCURATE AND SPEED TROPICAL RAINFORESTS CLASSIFICATION
Abstract
The accuracy of rainforests classification is generally improved by the input of multisensory data since complex vegetation type identification benefits from complementary information. However, in some cases, multisource fusion can also deteriorate accuracy when irrelevant sources are added. Thus, we introduce a fusion method for classes "in difficulty". Our method outperforms the classical global approach consisting in performing fusion for all classes. Moreover, the fusion processing time can significantly decrease when several classes are put aside. This operational method can be used effectively to enhance accuracy and processing speed when analyzing the wealth of information available from remote sensing products.