Multi-source Image target-region recognition based on DS theory

DS Theory, as one of the most important decision-level fusion algorithms, can combine decisions from different sources, so it has been widely used in the field of multi-source remote sensing images fusion. This paper first analyzes the basic knowledge about DS and its combination rule, and then for the 5 types of target-region in the JERS-1 SAR imagery and TM multispectral imagery about the neighborhood of Harbin, uses the grey feature of TM and the texture feature of SAR to recognize the target-region separately. After that, the correlation of target-region and region template is converted to the mass function probability and the unknown probability in DS. For the different influence on the target-region recognition of different sources, we can adjust the unknown probability of spatial information. By the DS combination rule, the two-source recognition results are combined. Finally we verify the effectiveness of the method by the priori knowledge. The experiments show that the combination result of DS can be more accurate than that of single source in the area of the multi-source target-region recognition. Based on the processing, we analysis some problems needed to be noticed in the DS application.