Data fusion is an effective tool to process multi-source remote sensing images. The data fusion framework and configuration comprising of pixels-level, feature-level and decision- level fusions in detail, and their relationships were introduced. The design and implementation methods of data fusion platform were analyzed, and DS evidence theory based and fussy Kolaonen neural network based cluster algorithm as a modi~ied testing method were selected. The result shows that the cluster accuracy and speed of fuzzy Kohonen neural network cluster algorithm is better than that of traditional algorithms.