| Geographic information digitalization is one of key tasks in Geographical Information System development, which takes about two thirds of the whole workload in GIS. Involving several subjects, automatic interpretation and vectorization of topographic map is a challenging technique with theoretical and practical values in digital image processing and recognition. It can improve the efficiency of digital maps productivity remarkably. Previous theories and applications have provided solid basis for further research, but there are still some urgent and thorny problems available. Firstly this dissertation analyzes recent developments in automatic map interpretation and vectorization, then problems in these researches are pointed out. With the direction of information extraction principles of human vision, several new thoughts and algorithms for automatic recognition and vectorization of the topographic map are proposed, such as the preprocessing of map image, color segmentation, contour lines vectorization, black elements recognition and dotted lines vectorization. At last, summing up these algorithms, a prototype system of automatic recognition and vectorization of topographic maps is developed. It proves the virtue of these algorithms and makes groundwork for a practical system at the same time.Main contributions of the dissertation include:l.A thorough research on the available methods and techniques of map information acquisition is performed. Based on the comprehensive analysis of the recent developments in this field, main problems in the research are pointed out.2. The color error in the digital image of topographic map is analyzed. With the summary and analysis of available anisotropic diffusion algorithms, a new adaptive diffusion algorithm based on image characteristics is presented. It overcomes the defects of the original algorithms such as sensitivity to noise, corner blurring. Furthermore, it performs robustly in other graphic images as well.3. Based on the analysis of the color error in a map image, the deficiency of available algorithms that only make use of color information in map segmentation is indicated. With the color space transformation and improved fuzzy c means clustering, the segmentation of topographic map is implemented. As a result, colorABSTRACTerror is restrained and the precision of segmentation is enhanced. This algorithm has established a fine basis for the recognition and automatic vectorization of color map.4. Based on the analysis of current vectorization algorithms and the characteristics of contour lines, a novel deformable model for automated tracking is explored. With the seed segment searching and variable forces control, contour lines are vectorized automatically. The tracking algorithm is significant because it is directed by adaptive field gradient flow with global analysis. At the same time, broken and cohering contour lines can be disposed automatically. Experiments prove that the model is practical and robust.5. Based on the morphological analysis of back elements in topographic map, a new effective algorithm based on morphological decomposition is presented. According to different morphological features, the structure of map elements is decomposed, and then elements of the same characteristics are extracted and combined through the algorithms of erosion, dilation, improved run-length smearing. As last, the elements are extracted correctly with global character analysis and the problems of overlapping and adherence among elements are resolved efficiently.6. The perceptual organization phenomenon in the cartography of topographic maps are analyzed and the significance and necessary of perceptual organization algorithms in interpretation of topographic maps is point out. Conforming with perceptual rules, a serial algorithms are put forward for automatic vectorization of dotted lines. Firstly, the image is filtered by mathematical morphology. The limited depth prior search is applied to generate the seed segment...
|
PDF Full Text Request