Improving road detection on SAR images using fuzzy fusion methods

This paper focuses on the use of fuzzy fusion techniques to improve the automatic detection of linear features on multi-temporal SAR images. Different fusion strategies involving different fusion operators are presented. Since T-norms and T-conorms do not lead to satisfactory results (these operators are respectively too severe and too indulgent), the first strategy consists in fusing the data using a compromise operator. The second strategy consists in fusing the results computed with two operators with opposite properties, in order to obtain a global intermediate result. Thanks to the wide range of behaviours they provide, fuzzy operators are used to test and compare these two fusion strategies on real ERS-1 data