In this paper, an information fusion method based on fuzzy set and evidence theory was proposed. Firstly, the fuzzy membership function of the recognition framework was obtained from the uncertain observation collected by the sensor. Then by using the membership function, the mass function could be obtained. After that, the evidence distance was used to determine the credibility and the discounting factor of each sensor. The present approach was tested in a target recognition system to illustrate its efficiency. The simulation results show that even if the evidences are conflict with each other in a high degree, correct results can be obtained.