Abstract: In this paper, after discussing existing three dimensional spatial data models, the characters of vector data model, raster data model and hybrid model are analyzed.An object-oriented integrative 3D data model is put forward.The model divides 3D space into raster voxels.All 3D objects are filled with raster voxels.Object-oriented method is adopted for object expression.For every object, contained voxels only are stored.Point, line, surface and solid are linked by hierarchy structure and space relation.Common point and line and surface are simultaneously stored in adjacent objects.In the aspect of space process, raster method is adopted for operations related with position.Vector method is adopted for calculation and query of space relation. The integrated model has the advantage of both vector and raster model.But it is different from raster model, vector model and hybrid model.Firstly, raster model connects voxels to expresses object through the attribute of voxels.But in the integrated model, raster elements have not attribute value.Voxels belonging to an object are orderly stored in a record.In other hand, raster model stores solid nodes in space order.The space index and expression of object relation depend on the attribute of voxels and reasoning by neighbor and connectivity.But the integrated model organizes sub-space of object by hierarchy structure.Secondly, in vector model, sub-space is divided irregularly and out-of-order.only expresses end point, boundary and surface of objects.But the integrated model fills and expresses interior spaces except for above elements.The difference from hybrid model is that hybrid model must simultaneously store raster and vector data.At the same time, transformation between raster and vector model is necessary, but sometimes it is difficult.In the integrated model, only raster data is stored, but the information about the position, shape, structure and topological relation, etc., can be described.A remarkable problem is that the numbers of data is enormous and the computing and query speed is slow for raster division to 3D space.In this paper, a method of three-grade division of 3D space is proposed.A rough division is carried out in a first grade.The second and third division is respectively carried out based on above division.The purpose of the third division is to improve the expressive precision of object, the division is carried in boundary of object, such as points, lines and surfaces.So the numbers of data is not remarkably increased.Subsequently, a row-order code method is proposed.The code method is superior to Morton code in query calculation and window display.The storage space in need is also small. Finally, the data structure of an integrated model is presented.