Research On Vectorizing Method For Classified Raster Data From Remote Sensing Imagery

With the development of Remote Sensing and Geographic Information Systems, remote-sensing (RS) imagery has become an increasingly important data source in GIS application. Vectorization for RS classified raster data is the inevitable way to use such kind of data in GIS spatial analysis. Thus, there is an urgent need to develop highly efficient and fast vectorizing algorithm for very large and complicated classified raster data. Therefore, the characteristics of RS classified raster data is systematically studied, and the vectorizing algorithms are proposed and implemented in this thesis.The primary contents of the thesis can be summarized as follows:(1) A window based raster data vectorizing algorithm is proposed. The fitteen kinds of data types in a2×2window are summarized for RS imagery classified raster data, and a window-based vectorizing algorithm is designed on that basis. All nodes and the coordinates of the points and the connection information between them are extracted by moving through the image using a2x2template window. Then, arcs are tracked by using connection information between all vertor points. Polygons also can be formed based on the connection information between nodes and arcs and topological relations between polygons are established. The practical experiment and comparative analysis with other algorithms indicates that the processing efficiency of the proposed algorithm is improved, and topological relationships can be formed and comparatively large image can also be handled.(2) A raster data vectorizing algorithm based on Run coding is proposed. The algorithm employs the correlations of adjacent grid points in raster data. The process of vectorization is executed polygon by polygon. The extraction of outer ring and inner rings for polygon can be performed together with building topological contain relationships. Experiments on real RS classified data are carried on to demonstrate that this algorithm can process the large raster data on a PC which has the amount of data over100M and the number of polygons more than one million and the efficiency of algorithm can meet the practical requirements. In addition, the algorithm create dynamically topological contains relationships while executing the extraction of outer rings and inner rings simultaneously. Thus, this algorithm can adapt to the case of complex island polygon.(3) The capability of processing the problem in the result of vectorization, such as complex island polygon, self-intersecting polygon, reducing the vector data, is comparised between the window-based and run-coding based vectorizing algorithm. Considering the large volume and complex characteristics of the Landuse RS imagery classified raster data, run coding-based vectorizing algorithm is chosen in the prototype. Finally the vectorizing and false-change removing prototype system is implemented using Visual C#2008, including spatial data reading and storing, batch vectorizing of classified change information (supporting large, complex raster data), and false-change removing, etc. functions.