Recognition And Acquisition Of Line-like Elements In Color Maps

Map vectorization is one of the key tasks in G1S development, which takes about two thirds of the whole workload to realize, so research on map vectorization means much to improve accuracy and efficiency of GIS development. Previous theories and applications provide solid basis for further research, so that we can focus on some urgent and thorny problems. This paper takes color map with a scale of 1:50000 as the research object and emphasizes on the recognition and acquisition of line-like elements.Main contributions of the dissertation include:1. Characteristics of map images are analyzed synthetically from their color, shape and structure. The influence of scanning process is introduced as well as a significant reason for difficulties in keeping information complete when color maps are transforme-d into mono-color maps and in solving map vectorization as a whole.2. A new procedure of map vectorization is proposed featuring carrying out thinning to all elements first and then segmenting them. According to previous research, segmentation, thinning and automatic track have become necessary techniques, but the new idea shows obvious advantages in simplifying objects by thinning, ensuring a good result in post-processing and on the other hand, relaxing the requirement to segmentation results.3. A practical segmentation algorithm is presented directly based on RGB color model, which saves computation for color space transformation, achieves similar results with other algorithms and adapts well to the variety of image quality and map version.4. The loss of geographical information is inevitable in segmented images, leaving many broken lines and points, which constitutes an important factor to bring down track accuracy and automatic level. In this paper, how broken points appear, how many categories they fall into and how to connect them by their structure features are all discussed in detail.5.On the basis of the above algorithms, a prototype system for recognition and acquisition of line-like elements in color maps is developed on PC and experiments are done to prove their effectiveness and efficiency.With practicability as the goal, this paper propose a complete procedure and algorithms for map vectorization are experimentally proved easy to realize, fast computation and adaptive to variety of image brightness and contrast, and information can be acquired in an accuracy of over 90%, thus greatly enhance efficiency and automatic level of map information acquisition.