A FRAMEWORK OF COLLABORATIVE CHANGE DETECTION WITH MULTIPLE OPERATORS AND MULTI-SOURCE REMOTE SENSING IMAGES

This paper proposes a framework of change detection with multi-source remote sensing images through collaboration of multiple operators. Firstly, pre-processed images are distributed to different operators. Then the images are classified by the operators independently. Finally, with uploaded classification results, change detection result can be derived through evidential fusion based on PCR5 rule in the server. By making use of complementary and redundant information in the images, the framework can solve the problems of information loss, imprecision, inconformity or conflict in multi-source data. The framework is applied to detect a landslide barrier lake with multi-source images from Landsat7 and GF-1, results show that as the amount of operator and input image increases, the proposed framework performs better than commonly used major voting strategy for disaster mapping.