Structure information is one of the most important types of auxiliary information in schema matching. When a schema has multiple elements with same semantics, the structure information is the most effective information to get correct matching for these elements. This is especially important in big-scale schema. Existing schema matching methods have some weaknesses such as inaccurate structure information, lack of effective description form, and high time complexity in the utilization of structure information therefore greatly hindering the use of structure information. In order to fully use the structure information, a new schema matching method based on information unit(IU_Based) is proposed in this paper. In IU_Based, the elements are first grouped in different information units according to the entity described. Then, the similarity between information units is calculated and the matching relation between information units is obtained based on the similarity. Finally, the matching between elements is selected from the matched information units. Extensive simulation experiments are conducted and the results show that IU_Based method can solve the problems in the use of structure information and take full advantage of structure information to improve the accuracy of match result.