In order to calculate the spacing and number of grid cell and get rid of data redundancy caused by subjective experience on the process of spatial partitioning, multi - factor coupled model of spatial partition was established by four constraints—Laves partition identity triangulated regular network, terrain features, data density and adaptive boundary. The model can establish adaptive boundary conditions based on discrete center and the furthest point of samples, and in order to reduce complexity and systematic errors of direct solution of terrain surface equations, the constraints of terrain features and data density were expressed by the terrain surface differential equations. The model's algorithm, based on recursive algorithm and automatic generation technology for discrete grids, can generate the 6 resolutions of grids and calculate the spacing and number of each resolution cell. Finally, when the sets of data from 110 to 440, by visualization and comparative analysis of 3 kinds of spatial partition methods, the results show that the cost of coupled model’s algorithm is 1/10~1/2 of the Delaunay and Square, and also verify that the algorithm is linear convergence and can effectively solve the problem of illegal border and polygons from non – convex.