Abstract: There are diverse kinds of sensors can be used for space objects recognition, such as radar, radio, IR sensors and electrooptical sensors. The information acquired by them is redundant and complemental in time domain, frequency domain or characteristic domain. The redundant information can improve the recognition reliability of fusion method and the complemental information can span the time or space range that the objects can be recognized. Various kinds of classifiers are designed to the complex space objects recognition problem, the automatic optimization and assemble of them is the key factor for the good performance of the whole fusion system. In this paper, the traits of each kind of sensors in the distributed heterogeneous sensor information process system are analyzed specially, and then a multilevel enhanced fusion scheme for space objects recognition is brought out to exploit latent abilities of various kinds of classifiers and the software design of systematical level is presented. The scheme assembles different sensors and fusion classifiers automatically so as to exploit the latent abilities of various kinds of classifiers and to improve the efficiency and robustness of the system.