Research On Linear Features Automatic Generalization Based On Empirical Mode Decomposition

As the status of geographic information system (GIS) continuously improves in modern society, the needs of geographical information are increasingly expanded and demand levels become higher and higher. Furthermore, more and more geographic phenomena are observed, understood and descripted in different resolution, different spatial scales according to a variety of requirements. Also, the variable-scale spatial data are analyzed, processed and expressed. The Internet and the popularity of mobile services have promoted the emergence and development of variable scale characteristic GIS. Therefore, the oldest and classical cartography field——the cartographic generalization is energetic again as well as faces new challenges, making cartographic generalization a hot area of research and cutting-edge issues in GIS area.The research on how to use computer-based map database so as to express map quickly and in multiple-scale is not only very significant and meaningful to maps for their information carriage and transmission in the new digital environment, but also makes the traditional cartographic generalization conform to the new requirements in the field of mapping. The data and information in GIS is expressed mainly through linear graphic elements, thus automatically linear features generalization is an important aspect of multi-scale expression of the geographic information. At present, the theory and methods that use spatial geometric constraints to linear features generalization have been widely used. However, not much attention has been paid to the geographic information generalization by filtering method, especially through empirical mode decomposition, which is an adaptive and natural approach for signal analysis. Empirical Mode Decomposition Method can well simplify and compress the linear elements. In addition, spatial geometric methods based on curvature have strong ability to maintain the shape of the characteristics curve in linear features generalization process. Therefore, the combination of spatial geometric constraints and filtering method is used for the automatic linear features generalization, and the main contents of this research include the following aspects:1) The existing problems of cartographic characteristics as well as the linear features generalization in digital environment has been analyzed, and the necessity of using frequency domain filtering method in linear features generalization is discussed. By analyzing and summarizing the main problems in line generalization methods, the technical route, which uses curve empirical mode decomposition in two-dimensional space, is proposed to compress and simplify the linear elements.2) A variety of spatial information generalization methods using filtering theory are studied. The empirical mode decomposition method is put forward based on summarizing and analyzing the theory of wavelet, Fourier transform and empirical mode decomposition theory. Through the empirical mode decomposition of lines in the two-dimensional space, multi-scale representation of linear features method is brought forward by using Bessel and BSpline curve approximation method.3) The concept of visual curvature has been introduced, what is more, the advantages and disadvantages of curvature method in curve feature point detection are probed. Secondly, by summarizing the previous methods of curvature calculation, visual curvature computation based on elevating angular height calculation is proposed. Through experiments of polygon shape extraction based on visual curvature, curvature-based visualization method for curve simplifying is adopted.4) The empirical mode decomposition algorithm for two-dimensional space and the curve reconstruction algorithm based on constraint features have been designed by combining empirical mode decomposition and curvature visualization method. By adopting the constraints such as spatial semantic relationship, the method for preserving the shape characteristics of linear features after generalization is establised.5) The analysis and evaluation on results of linear features generalization is carried out, and an improved Hausdorff similarity curve evaluation method is proposed to examine the shape of curve characteristics after generalization. At the same time, both geometric characteristics assessment method and location accuracy evaluation method are used to analyze, evaluate the linear features generalization outcome. At last, the characteristics of simplifying linear features based on EMD have been pointed out. Also, the applicability of two-dimensional curves empirical mode decomposition in linear features generalization is indicated.