A Map Recognition Method With Self-Diagnostics Function And Its Implements

This article is concerned with the study of map recognition, and in the article much emphasis has been put on the innovation of researches in the past has either been focused on individual aspects (e.g vectorization) of map or taken sorts of information-losing techniques in accordance with bottom-to-top principle. Such a mode of researching slows down the development of map recognition study. This article take a map as complex system and takes an overall view on map recognition study according to the ideology of system methods. Not only are map elements are studied, but also the relationship between the structure and elements of a map system, and further study are conducted on such a basis. While making studies on the characteristics of map systems and map recognition algorithms, this article also puts much emphasis on the organic connections between those algorithms and the realization of self diagnostic function. This self-diagnostic function is implemented, not simply by a bottom-to-top basis, but by human vision. A combined control model with self-diagnostic feature has been proffered.The combined model with self-diagnostic feature is composed of the following functions: separation and extraction of map segments and nodes; contour extraction of map segment and node, vectorization of contour and skeletonization of map segment; the building of adjacency relations of map segments and nodes; secondary recognition of map segment and map element according to restrictions; recognition of map segment vectors and map elements.Concerning the separation and extraction of map segments and nods, this article proposed the notion of map segments and nodes, and a algorithm of run growth method to separate map segments and nodes, through which excessive local processing can be avoided and overall map features can be grasped. This relieves the contradictions between overall and local methods of recognition in certain degree.Concerning the skeletonization of map segments, this article has described the whole process of the extraction and tracing of contour points of map segment, vetorization of contours and skeletons of map segment. A one-off method that extracts and traces contours points simultaneously has been proposed. Aiming at solving several problems in the contour approaching processes, a liner approach method based on Least Square Window Error Linear approach method and linear combination principle. The method is suitable for different circumstance of recognition.Concerning problem solving with restrictions, this article propose that itinerant the searches with restrictions be carried out in different levels of information. As to implementation, two methods based on Hough transform and Least Square Error has been made.Concerning recognition, this article proposed a method that combines labeling and matching. The self-diagnostic function is carried out in two steps,(1)validation of map segment vectors, i.e. verify that a map segment and its corresponding segment in raster image are consistent with each other and satisfy appropriate restrictions. Go ahead to next stage of accurate solving if they are consistent or else, report that the segment dissatisfy the restrictions, and redo the step, (2) extract main features of the map element obtained, and make further diagnosis on it according to the restrictions it should satisfy, end the diagnosis if it satisfies the restrictions and report a problem and re-recognize if it does not.Finally, this article introduces a map recognition system implemented by the author. The structure and function of the system as well as several other techniques pertaining to map recognition has been introduced.