SOM神经网络在道路网制图综合中的应用

发布时间：2018-05-03 21:32

本文选题：SOM神经网络 + 道路网　；参考：《南京大学》2013年硕士论文

【摘要】：在地图制图中,因为地图图幅有限,地图上不可能将研究区内所有的要素表示出来,因此只能根据地图用途、地图比例尺和制图区域特点,将制图对象中的规律性和典型特征以概括和抽象的形式表示出来,而对于那些次要的和非本质的地物要素则要舍弃。当地图由大比例尺缩小到小比例尺时,地图的面积是按等比级数缩小的,如果将地图上的要素也按照这种方式缩小的话,地物要素将出现扭曲、压盖、变形等现象。为了解决地图要素与实地地物之间的矛盾,则需要用到制图综合,它通过对地图内容的选取、化简、概括和位移,可以建立反映区域地理规律和特点的新的地图模型。传统的制图综合主要是靠手工实现,这种制图方式要求繁琐的手工劳动,而且存在很大的主观性。此外,手工制图综合的质量也会受到很多人为因素的影响,所以很难保证地图的质量和品质。现代制图综合指的是计算机环境下的制图综合,数字制图技术在很大程度上促进了地图生产效率的提高,为地图自动制图综合的研究提供了技术基础。在数字制图技术条件下,一方面地图自动制图综合将制图人员从繁杂的手工操作中解脱出来,另一方面制图人员可以投入更多的精力研究如何提高制图综合的自动化程度。在计算机技术和图形交互编辑功能不断提高的背景下,自动制图综合的理论和应用也取得了很大的进步,但仍然存在很多难以克服的问题,比较突出的问题是制图综合的自动综合程度还不是很高。人工智能技术在自动制图综合领域中的应用是该领域的一个重要突破,因为人工智能技术具有一定的人脑思维能力,从而可以在某种程度上模拟人类制图综合的过程。人工神经网络是一种采用物理可实现系统来模拟人脑神经细胞的结构和功能的系统,它具有自学习、联想存储、和高速寻找优化解的功能,因此可以把它应用于制图综合中。线状要素的制图综合是制图综合领域中的研究热点和重点。道路网遍布全图,形状多样、关系复杂、等级繁多,是所有地图要素中比较重要、使用频率较高的数据层,有着重要的经济和军事意义。因此,使用人工神经网络研究道路网的制图综合具有重要的意义。本研究使用一种方法,它将道路的拓扑、几何和语义属性输入到一个自组织竞争神经网络中,自组织竞争神经网络是一种人工神经网络,在此研究中用于对道路网的聚类分析。更具体地说,该方法根据多种属性将所有的道路分成不同的类,然后基于这些分类在比例尺缩小的地图上按照某种指标对道路进行选取。传统的道路选取方法主要是根据道路等级等语义属性进行选取的,而忽视了道路的空间特性,本文分别使用了道路的拓扑、几何和语义属性将道路进行聚类,考虑比较全面,因此聚类结果更加准确,在此基础上对道路进行选取也可以得到更好的效果。
[Abstract]:In cartography, it is not possible to express all the elements of the study area on a map because of its limited map size, and therefore can only be based on the purpose of the map, the map scale and the characteristics of the cartographic area, The regularity and typical features of the mapping object are expressed in the form of generalization and abstraction, but the secondary and non-essential elements of the feature should be abandoned. When the map is reduced from a large scale to a small scale, the area of the map is reduced according to the equal-ratio series. If the elements on the map are also reduced in this way, the elements of the feature will be distorted, overlaid, deformed and so on. In order to solve the contradiction between map elements and field features, cartographic generalization is needed. Through the selection, simplification, generalization and displacement of map contents, a new map model can be established to reflect the laws and characteristics of regional geography. Traditional cartographic generalization is mainly realized by hand, which requires complicated manual work and has great subjectivity. In addition, the quality of manual cartography generalization is also affected by many man-made factors, so it is difficult to ensure the quality and quality of maps. Modern cartographic generalization refers to cartographic generalization in computer environment. Digital cartography technology promotes the efficiency of map production to a great extent and provides a technical basis for the research of automatic cartographic generalization. Under the condition of digital cartography technology, on the one hand, automatic cartographic generalization can free cartographers from complicated manual operation, on the other hand, cartographers can devote more energy to study how to improve the automation of cartographic generalization. With the continuous improvement of computer technology and interactive editing of graphics, great progress has been made in the theory and application of automatic cartographic generalization, but there are still many insurmountable problems. The outstanding problem is that the degree of automatic generalization of cartographic generalization is not very high. The application of artificial intelligence technology in the field of automatic cartographic generalization is an important breakthrough in this field, because artificial intelligence technology has certain human brain thinking ability, so it can simulate the process of human cartographic generalization to some extent. Artificial neural network (Ann) is a physical system which can simulate the structure and function of human neural cells. It has the functions of self-learning, associative storage and high-speed searching for optimal solution, so it can be applied to cartographic generalization. Cartographic generalization of linear elements is the focus of research in the field of cartographic generalization. Road network is the most important data layer in all the map elements and has important economic and military significance. Therefore, it is of great significance to use artificial neural network to study road network cartographic generalization. In this study, a method is used to input the topology, geometry and semantic attributes of the road into a self-organizing competitive neural network, which is an artificial neural network, which is used in the clustering analysis of the road network. More specifically, the method divides all the roads into different classes according to various attributes, and then selects the roads according to a certain index on a scaled down map based on these classifications. The traditional road selection method is mainly based on the semantic attributes such as road grade, but neglects the spatial characteristics of the road. In this paper, the road topology, geometry and semantic attributes are used to cluster the road. Therefore, the clustering results are more accurate, and better results can be obtained on the basis of the selection of roads.
【学位授予单位】：南京大学
【学位级别】：硕士
【学位授予年份】：2013
【分类号】：P283.7

【相似文献】