The flow field variation and fish attraction effects in artificial reefs are usually realized through different scales of forms such as unit fish reef and fish reef cluster.What distribution status of the hundreds and thousands of artificial reefs actually launched is and if such reefs meet the requirement of unit reef worth very much attention.So,how to conduct reasonable classify or eliminate reefs and take the maximum extent to fit design scheme,is very necessary and has practical significance for evaluating whether ARs are launched accurately.In this paper,based on high-definition underwater images collected by C3D side-scan sonar system,combining with ArcGIS,we can extract the submarine space position of each fish reef monocase as well as spatial relationships i.e.mutual distance and azimuth between them etc.On this basis,a trial on spatial cluster analysis based on partition,hierarchical and constrained algorithms,was compared in order to find a relatively suitable solution for interpretation of distribution.Centre of gravity for deployed reefs,their potential influenced area,number of unit reefs and distance between each of the two reefs are defined as four basic indices for further comparison.Results show that the error of three spatial clustering algorithms in an ascending order is:constrained algorithm< partition algorithm< hierarchical algorithm,the values were 0.093,0.203,0.264.Comparative analysis shows that the constrained cluster algorithm can reflect the actual distribution pattern of the artificial reefs in a best way.