The Study Of Auto-vectorization Method Based On Scanning-Thinning Method

From the 1960s, since the rapidly developing of the Cartography, GIS and RS technique promoted the development of the Mapping technology. It is important modern way of getting geographic information that the vectorisation of existing papery maps and the classification of RS images. Since the construction of the Topographic Map Database (1:10000) in China the vectorisation technology is more and more important. The Raster Map includes the Scanning map, the Classification diagram of RS Image and Raster analysis result of Raster map. Now there were three methods of vectorisation:Manual Vectorisation, Semi-Automatic Vectorisation, Automatic Vectorisation (and the post-processing). The first method is slow, precision varied with the technology and attention of the drafters, and less errors; The second method with high precision but speed is general, and lower errors; The last method is high speed, high precision but need to reduce errors with post-treatment, and these algorithms obtained long time to post-processing, sometimes the time even more than the time of first method processing, so it's urgently needs to develop improvement.This paper analyze the method, the development and the research situation of Vectorisation of Raster Map; And point out the existing problems and the insufficiency of the existing Automatic Vectorisation, and something in need of improvement. This paper analyze the basic theory which related the; And its sums up and improved some processing methods, such as:the Cubic is the best Re-sampling methods for Raster map, and the pixel size should selected from 0.1-0.4 times wide of the thinnest line in the map; The Catch-Grids-Ajustment; And Based on color to track the different classes of geographical elements, etc.The main subject of this paper is Raster-line-thinning method and automatic-tracking method, and makes a detailed analysis and research on the two subjects. By designing, experiments, updating algorithm, and finally this paper put forward the Scanning-Thinning-Algorithm based on the morphology of raster lines. This algorithm have small influence with the boundary of raster line, and separate the result pixel with two classes:ordinary pixel and characteristic pixel, the first class for lines' connected properties; And the second class, which include convex pixel, feature pixel and cross pixel, used to convert to vector coordinate, and greatly reduced the amount of compute and the storage space; And improved the accuracy. In addition, the result of Scan-Thinning-Algorithm difference from traditional not only with the format but also with the function. We are design an algorithm for auto-tracking the line from raster to vector to adapt the result of Scan-Thinning-Algorithm, and make it. In tracking process, we takes the Vector-compressing on the time based on the "The compress method of three distance with five point".Finally the paper slightly analysis the automatic methods and post-processing of vectorisation of the grayscale linear map, colored linear map and colored area map; And make experiment to proved that the feasibility of the algorithm and the method in this paper.