Abstract: In order to evaluate the overall performance of data fusion system, an approximately complete set of criterions is established for the distributed kind. An evaluation model of grey relation analysis based on rough set theory is proposed, which by means of applying rough set theory reduce redundant attributes and determine criterion weight values, thus resulting in less subjectivity and more objective evaluation results. The feasibility and credibility of the model is proved by an example. It provides a scientific decision foundation theoretically for development, demonstration and practical application of the data fusion system.