Abstract: Nowadays, the existing methods usually utilize the Voronoi diagram for automatically generalizing the islands in chart, but it is hard for them to make the generalized results meet the actual requirement due to the significant reasons that that the Voronoi diagram cannot represent the density and distribution of the islands accurately in chart cartography generalization. Aiming at these drawbacks, a method for islands automatic selection based on buffer dynamic growing model is proposed. Firstly, the drawbacks of Voronoi diagram representing density and distribution of spatial data are analyzed theoretically, and the buffer growing model is designed which can overcome these drawbacks. Secondly, the concepts of distance matrix, island buffer's identification and growth factor are defined, and the buffer growth model is constructed. Then the mechanism for updating the model dynamically is designed. Finally, the islands in chart are automatically selected through building and maintaining the model dynamically, and several groups of islands with different types of sea areas are selected for experimental verification. The experimental results show that:(1) the buffer growth model is effective for representing the density and distribution of the islands; (2) the proposed method improves the quality of the island which is obviously much higher than that of the methods based on Voronoi diagram, and it can be applied into more types of sea areas.