Research And Implement Of Point Symbols Recognition And Text Extraction In Color Topographic Map

Map pattern recognition is the core technology to improve the efficiency of map automatic digitization. As two important parts of map pattern recognition, point symbols recognition and text extraction in color topographic map are the emphasis and difficulty in present study. It is of great significance for the development of map automatic digitization to improve the accuracy of point symbols recognition and text extraction.This thesis demonstrates the research significance and the current development situation of map point symbols recognition and text extraction, and analyzes the characteristics of point symbol and text in color topographic map. The two aspects—point symbols recognition and text extraction, are mainly studied in this thesis, and two algorithms—Generalized Hough Transform with Matching Feedback and Text Extraction using Character Size and Connection Curvature are proposed. In addition, two components are developed for actual collecting in the Map GIS K9 platform.In the term of point symbols recognition, for the issue that point symbols can't be recognized using the traditionally method that point symbols must be extracted firstly because of that the point symbol adheres, crosses and even covers with other geographic information elements. A new algorithm, which is called Generalized Hough Transform with Matching Feedback to recognize point symbols, is proposed in this thesis. On the basis of traditional Generalized Hough Transform algorithm, color information is introduced to MP-GHT so this algorithm can make the best use of the features of color and shape to recognize point symbols in color topographic map directly. In addition, fuzzy matching is added after the preliminary recognition, so the result of the preliminary recognition can be further modified by the matching feedback to improve the accuracy of point symbols recognition.In the term of text extraction, a new algorithm called Text Extraction using Character Size and Connection Curvature is proposed in this thesis, according to the size of characters, the connection curvature of characters, and the characteristics of the region that character is adhered even crossed by lines in the color topographic map. First, using the color information of text in the color topographic map, the image that contains text is separated by FCM clustering color segmentation and the connect component that doesn't belong to characters obviously are deleted. Then, words are pre-grouped using the size of character and the connection curvature of characters, and according to the relation between connect components in the same word, post-process is performed to improve the accuracy of text grouping. Finally, a method called rotation-corrosion is used to orientate the text that the oblique text is corrected into horizontal direction, so a good input can be provided to the text recognition subsequently.According to the topic research achievement on point symbols recognition and text extraction, this thesis develops two components—point symbols recognition component and text extraction component in color topographic map respectively in the Map GIS K9 platform on the basis of COM component technology standards with Open CV open-source computer vision library and Visual Studio 2005 for the actual collection of geographical elements work.Point symbols recognition and text extraction in color topographic map are mainly studied in this thesis. The accuracy of two algorithms—Generalized Hough Transform with Matching Feedback and Text Extraction using Character Size and Connection Curvature are more than 90%. In addition, two components called point symbols recognition component and text extraction component in color topographic map respectively in the Map GIS K9 platform have been used in actual collection of geographical elements.