Dissertation > Excellent graduate degree dissertation topics show

Intelligent Recognition of Line in Scanned Color Maps Based on Segment

Author: LiHuaRong
Tutor: DuQingYun
School: Wuhan University
Course: Cartography and Geographic Information Systems
Keywords: Map Recognition Scan string Figure segment connected component Line symbol Contour Road River
CLC: TP391.41
Type: PhD thesis
Year: 2010
Downloads: 83
Quote: 0
Read: Download Dissertation

Abstract

An important part of geographic information system (GIS) geographic data acquisition, more than 80% of the work accounts for the entire system development, but also affect the bottleneck in the development of GIS. The map features intelligent identification is the core issue of geographic data acquisition, image processing, pattern recognition and artificial intelligence, multi-disciplinary integrated application, but also an important issue of the field of computer vision. Its direct-to-business needs, with high theoretical significance and application value, years of theoretical study and practice the in-depth study of this subject has laid a good foundation, but there are many problems to be solved, recognition theory and methods to be breakthrough. Topographic map organic collection of point, line, surface, symbol, vector DLG; scanned topographic map image dot matrix pixel natural collection. Dot pixels polymerization map features symbols from the knowledge, reasoning hierarchical processing and self-organization, seeks to strengthen the integrity of the expression unit, to improve the level of expression, focus on a variety of related information and heuristic information selection and organizations to identify the data. Extended FIG segment constraint conditions, so as to satisfy the requirements of the color, width, and topological consistency, Fig segment image expression unit having a single meaning. According to the association between the pixels, the dot matrix image first run-length coding, building scan string with attribute information, and then formed based on the the scan string width, color, and the topological consistency Figure segment, image map segment expression, while extraction Figure connection relationship between the segment, to build Fig segment connected component. According to the the Figure paragraph connected component attribute extraction point, line, surface connected component and line vectorization, get the connection line between the symbols and their relationship to build the adjacency relationship between the line symbol. According to the correlation between the line symbol, search all belong to the same map elements line symbol, the relationship between the extraction, reconstruction of map features. This paper presents three models of topographic map - map element model based on the topographic maps of the human brain to identify the target information organization, symbolic model and image models, symbols linked model realistic model and image model, the model is based on integration of vector and raster data representation form, feature-oriented features with vector data; same time, by the identification number of the feature can be found in the feature means all the pixels, which in turn has all the features of the grid. Feature identification and extraction, can take advantage of the feature pixel local features, while at the same time the overall characteristic feature topological relations based on symbolic model. The extracted information into the recognition process of FIG segment, symbols, and three levels of map elements, the same level of data between interrelated data between different levels are interrelated, but also to the longitudinal direction Relevance Relevance both transversely oriented. In recognition, the first pixel local feature to generate Figure segment connected component of the overall structure, and then to guide the overall structure and correction of local features; senior information obtained from the low-level data, which in turn went to the guidance of the low-level data, the use of bottom- reasoning combined with top-down manner to complete the entire identification. In order to take full advantage of the color information of the topographic map, based on the analysis of the previous color processing method, the idea of ??increase the number of color classification for the pixel level, which effectively solve the problem of fuzzy transition color separation; Scan-oriented string level color combinations grouped into 16 species, with the color code and subtly, so that the scan string with color information, and to provide a basis for Figure-color information extraction. In this paper, the run-length coding technology to achieve a like yuan matrix → scan string → Fig segment conversion. FIG segment is to satisfy the color, width, and topological consistency adjacent scan string, which can direct expression of the line segment and the point of intersection. Figure of the color information and adjacency to build a single layout segment connected component, in order to achieve automatic color separation of the topographic image, taking into account the color attribute and spatial relationships, the more superior in terms of processing efficiency and noise suppression. Single version of the connected component graph based on graph segment connected component analysis. Start from the the map symbol shape, size and topological relations summarized suitable for expression of the characteristics of the feature size, aspect ratio, black-and-white than the node density and based on the characteristics of the difference between the line symbol from the point surface symbol separated. Extracted line symbol, according to the node graph segment using sub-vector approach, then the node graph segment adjacency homologous line detection, which all belong to the same map elements vector segment, form a complete vector information . Adaptively detection conditions for both the type of map elements, better extraction of the dashed roads, rivers, and the vector data of the contour lines. Based on the identification methods and algorithms, design and development of a prototype system of intelligent recognition of a topographic map. Software VC-platform, object-oriented design methods and the use of general-purpose database management data.

Related Dissertations

CLC: > Industrial Technology > Automation technology,computer technology > Computing technology,computer technology > Computer applications > Information processing (information processing) > Pattern Recognition and devices > Image recognition device
© 2012 www.DissertationTopic.Net Mobile