Computational Models And Methods For Orientation Of Buildings And Their Directional Relations

Buildings are such places for housing or working that they are closely related to people's daily life or other activities. In the fields of cartography and GIS, buildings are a class of very common thematic features in topography maps. Spatial relations among these features are in essence the spatial constraints for their locations, so they are very useful for spatial query, spatial analysis and reasoning. Spatial relations are often categorized into topological, directional and distance. This paper will be focused on the representations of both the orientation of building and the directional relations between buildings.Spatial orientation is defined as a kind of location or position information relative to the points of the compass within the given reference system, while the directional relation between two spatial objects is involved to assign reference object, target object and reference system. Indeed, Orientation of buildings and their directional relations play an important role in many aspects such as urban layout, mapping integrating, road building, and in GIS spatial query and spatial analysis. In practical applications, one of main problems is how to compute and represent the building orientation and the directional relations between buildings. To solve such problem, the paper pays much attention to developing computational models and methods for the building orientation and the directional relations between buildings.Indeed, orientation of buildings is one important spatial constraint in spatial data handling. It may be used to evaluate the quality of generalization algorithm. However, there is still not a sound method used to calculate building orientation so far. For this purpose, this paper firstly analyzes the five common methods, i.e. the longest edge, weighted bisector, statistical weighting, the smallest minimum bounding rectangle and wall average. Particularly, their applicability and limitations are analyzed according to their basic principles. Through two experiments, the applicability of these methods and their problems in use are further illustrated. It is also found that the smallest MBR is more robust than others. Furthermore, in order to calculate orientation of building groups, this paper tries to deal with buildings groups using mathematical morphology. And this method needs to match the given templets after processing images, and the orientation of building groups is determined depending on the natching result. This method has already been experimentally verified in the paper.Subsequently, this paper discusses the computational models of the directional relations between buildings. Firstly, the existing computational models of direction relations are reviewed in detail, including centroid-based model, cone-shaped model, projection-based model, four semi-infinite area model, direction-relation matrix model, 2-D String model, description model of detailed direction relations, statistical model. It has been found thatt most existing models are only approximate capture of the directional relations between extended spatial objects (i.e. lines and areas). In other words, they are just approximative models. In order to overcome this problem, we propose a new model based on spatial objects themselves, and called a statistical model. Through two questionnaire, the paper exemplify that the result of using statistical models is very close to human spatial cognition.Finally, the paper summarizes the findings, and some issues are highlighted for further investigations.