基于SOM的农业时空数据可视化分析模式

发布时间：2018-07-13 10:05

【摘要】：随着科学技术的迅猛发展,大量不同地理数据的收集,使得地理信息系统存储、处理和分析地理数据的作用日渐突出。当前的地理信息系统大部分处理的主要是静态的二维的数字地图。针对复杂的时空数据,用户更需要一个可视化的、交互的环境来处理、分析、显示数据。本文的研究内容主要包括:(1)选用SOM神经网络作为时空数据可视化分析的方法,基于SOM神经网络是一种针对高维数据聚类的神经网络,它不仅能够保持拓扑结构的不变,还能对复杂高维数据进行抽象概括后转换后在低维显示;(2)通过协同多视图的方法建立时空数据可视化分析模式,从多个角度对时空数据进行可视化表达与可视化分析;(3)时空立方体的可视化表达分析解译时空数据的时空变化。本文以美国NASS网站中农业调查数据为数据源,建立农业时空数据库。以Visio studio 2010为开发平台,采用C#语言编程,以农产品时空数据为例构建时空数据可视化分析模式界面。该界面中,以协同多视图的方法表达时空数据在不同组合下的分类及其演化过程,用户自由选择的属性组合和时间序列,系统以SOM神经网络进行时空数据的可视化分析,并以蜂窝图等可视化方式展示各类型的数据标签,距离,数量等,并最终以时空立方体的方法展示其时空分布和演变。
[Abstract]:With the rapid development of science and technology, a large number of different geographic data collection, making GIS storage, processing and analysis of geographical data increasingly prominent role. Most of the current GIS deals with static two-dimensional digital maps. For complex spatiotemporal data, users need a visual, interactive environment to process, analyze and display data. The main contents of this paper are as follows: (1) SOM neural network is used as the method of spatiotemporal data visualization analysis. SOM neural network is a kind of neural network for high dimensional data clustering, which can not only keep the topological structure unchanged. It can also be used to abstract and generalize the complex high-dimensional data and then transform it into low-dimensional display. (2) the visualization analysis model of spatio-temporal data can be established by collaborative multi-view method. Visualizing and analyzing spatiotemporal data from many angles. (3) analyzing and interpreting the spatio-temporal variation of spatio-temporal data by visualized representation of spatio-temporal cubes. Based on the agricultural survey data from NASS website, this paper establishes a spatiotemporal database of agriculture. Taking Visio studio 2010 as the development platform, using C # programming language and taking the spatiotemporal data of agricultural products as an example, the visual analysis mode interface of spatio-temporal data is constructed. In this interface, the classification and evolution process of spatio-temporal data under different combinations are expressed by cooperative multi-view method. The users freely select attribute combinations and time series. The system uses SOM neural network to visualize the spatio-temporal data. Finally, the spatial and temporal distribution and evolution of each type of data label, distance, quantity and so on are displayed by the method of spatiotemporal cube.
【学位授予单位】：浙江农林大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：S126

【相似文献】