Research On Automatic Extraction And Recognition Of Color Topographic Map Point Symbol

Geographic special data is a crucial part in CyberCity construction. Converting the old maps to digital format and extracting maps'essentials, resulting in achieving geographic information, is the one of the economic and effective ways. And it will be a main source of digital geographic information for a long while from now on.This thesis researched the automatic extraction and recognition topographic point symbol in detail. At first, it analyzed the process of human reading maps, and then using mode recognition theory, it also analyzed the process of computer reading maps. Based on above two efforts, the thesis designed the whole frame of map recognition. In order to improve the color map process result, it employed 2-step noise immunity model.After map scanning, the image efficiency can be improved by smoothing the color by the means of no-linear way and different characters in different direction. Comparison of various color space are made. Thus I1-I2-I3 color space is chosen as the appropriate one suitable for color space segmentation. The fussy c means clustering automatically segment the color topographic maps, generating four color essential elements such as black, blown, blue and green. Formatting these essential elements into two values (0 or 1), mathematic morphology is implemented to decrease noise, which benefit the following image recognition. The processed elements are vectorized after thinning. The statistic method and artificial neural networks are employed to recognize the point topographic symbol in black element map.