Abstract: Spatial clustering has played an important role in spatial data mining and spatial analysis. It aims to classify the spatial entities in a database into some meaningful clusters, where entities are similar to one another in same clusters and are dissimilar to the entities in different clusters. In practical applications, spatial clustering algorithm can be used to discover clusters in the spatial database with complicated structures, such as clusters of uneven density, clusters of arbitrary shape, clusters adjacent to each other, and clusters with significant outliers. In the meantime, it is also required that entities in a same cluster should be both nearest in spatial domain and very similar in thematic attribute. For this purpose, a novel multi-constraints based spatial clustering algorithm is proposed in this paper. Delaunay triangulation is first employed to determine the spatial proximate relationships among entities, and then multi-levels constraints are utilized to discover clusters in the spatial database. The presented algorithm is tested on both simulated and practical spatial databases, and its advantages over some well-known algorithms are also fully highlighted.