Scale-Space Theory Based Regionalization for Spatial Cells
多尺度空间单元区域划分方法

Traditional regional partitioning model for spatial cells is only built upon the information of attributes in every cell. However, the spatial relationships and their spatial interaction between cells are not considered sufficiently. In this study, based on scale-space theory a new approach for regional partitioning or regionalization for spatial cells is proposed so that the elements of spatial relationship between cells could be integrated besides considering the information of attributes. By this approach, regional partitioning in multiple scale can be accomplished on the basis of the spatial clustering algorithm that at certain scale the spatial cells could be melted into one class if their connective direction is the same within the road transformation system that is built upon by a spatial interactive model. Finally, according to the social and economic statistical data of 18 years from 1978 to 1995, the experiments of regional partitioning for social and economic development level of Jiangsu Province are achieved. In the experiments, despite only the simple spatial correlative model is used as the spatial interactive model for the scale-space clustering algorithm the regional partitioning results are highly accordant with real situations.