Research On Spatial Clustering In The Presence Of Obstacles

Spatial clustering is a primary field in spatial data mining with the task of grouping the objects of a spatial database based on distance into meaningful subclasses so that the members of a cluster are as similar as possible whereas the members of diferent clusters difer as much as possible from each other. Spatial clustering has many applicationsin GIS, such as remotesense, image processing, environment research and so on.The spatial clustering algorithms in the presence of constraints, specially the presence of obstacles, are introduced detailedly in this paper. Based on learning the current spatial clustering algorithms in the presence of obstacles, such as COD_CLARANS,DBCLuC,AUTOCLUST+ and DBRS+, a new method of density-based spatial clustering called COD_DBCLuC is proposed which can handle the obstacle constraints in a new way. In COD_DBCLuC, we use Obstructed distance to replace Euclidean distance in DBcluC as the criterion and propose a Polygon Combination and Reduction method in the pre-processing stage to improve the efficiency.COD_DBCLuC has been realized by C# program in this paper. Simulations based on datasets and the results showed that this new proposed approach has more effectiveness than DBCLuC and DBRS+ because it not only has the advantages of density-based clustering algorithms, but also takes advantage of the Obstructed distance to make the results more reasonable than traditional ways.