Research On The Self-Organizing Map’s Application In The Cartographic Generalization Of Road Networks

The map can not present all the object elements in the study area during the course of cartography since the map area is limited. Thus the solution is represent the object elements’s regularity and typical characteristics in the general and abstract way and abandon those less important and non-essential elements.The map area will be reduced by geometric progression when the map scale is reduced from big to small. Under this condition if the map elements are also reduced in this way, there will exist extortion,overlay and deform among the map elements.To solve the contradiction between the map elements and the real objects,map generalization is needed in the process, which relies on selecting,simplifying,generalizing and displacing the map elements to construct the map model which can represent the regional geography law and characteristics.The traditional cartography generalization is mainly achieved by hand,which is very laborious and has great subjectivity. In addition, the quality of manual cartography generalization will also impacted by many human factors,thus it is very hard to ensure the maps’quality. The modern cartography generalization refers to the cartography generalization under under the context of computer technology.The digital cartography technology has improved the map production efficiency to a great extent, which provides the technical foundation for the research of automatic cartography generalization. With the help of digital cartography technology, the automatic cartography generalization can release the cartographer from the tedious manual operations. Besides, it allows the cartographers to put more effort in studying how to improve the automation level of the cartography generalization.The computer technology and interactive graphics editing functions are continuing improving, under which background the theories and applications of automatic cartography generalization have achieved much progress. But there still exist many hard problems to overcome and one prominent problem is that the automatic generalization level of cartography generalization is not very high. The application of artificial intelligence technology in the field of automatic cartography generalization is one big breakthrough since this technology has a certain human thinking ability, therefore it can simulate the process of human cartography generalization. The artificial neural networks is a system using the physical system that can be realized to simulate the human neural cells’structures and functions. It has the ability of self-learning,associative storage and high-speed of finding optimal solutions, therefore it can be used in the cartography generalization.The cartographic generalization of line elements is the research focus and emphasis in the field of cartographic generalization. Road networks spread all over the map, and their shapes are diverse, relationships are complex and grades are many, thus they are the data which are more important among all the map elements and have been used frequently. Besides, they have the important economic and military significance. Therefore, it’s very important to use artificial neural networks to study road networks’cartographic generalization.This study uses a method to input the roads’ topology, geometric and semantic attributes to a self-organizing competitive neural networks. The networks is one kind of artificial neural networks,which is used for cluster analysis of road networks. More specifically,this method is based on many attributes to classify all the roads, and then relies on those classifications to select roads on a scale-reduced map according to a certain indicator.The traditional road selection method is mainly based on the sematic attributes like road classifications to select roads, while it overlooks the roads’ spatial attributes. This paper respectively uses roads’ topology, geometric and sematic attributes to classify the roads, which are more comprehensive than traditional method, thus the classification result is more accurate. Moreover, the result of roads selection is more satisfactory.