Proteins interact with other molecules through binding sites, so it is significant to identify protein binding sites. Although there are different computational methods for the identification of binding sites, the existing prediction methods have problems of low hit rate or large computation. In this paper, a binding sites prediction method based on structural alignment is introduced. In the process of structural alignment, homologous index is applied to screening out homologous templates, with which query chains are aligned, and then the ligands in similar structure templates are mapped onto the query chains. Clustering method is used for analysis of sites. The result indicates that reduced computation and improved prediction accuracy, compared with other prediction methods, can be obtained through our method.