【作者】 虞娟；

【导师】 陈一梅；

【作者基本信息】 东南大学 ， 道路与铁道工程， 2005， 硕士

【摘要】 本文利用实测地形图和遥感图像提供的岸线信息、水文统计数据等资料,以闽江竹岐至侯官段岸线为例对岸线演变进行研究。主要内容是:遥感岸线信息提取。选择合适的遥感图像进行图像恢复和增强处理。运用岸线遥感原理,采用二值法对遥感岸线数据进行提取,并对遥感岸线误差原因进行分析。通过遥感岸线精度分析,发现采用遥感技术获取的岸线误差较小,可用于岸线演变分析。利用闽江竹岐至侯官河段实测岸线资料和遥感岸线资料,对岸线动态变化进行了分析研究。所得成果揭示了闽江竹歧至侯官河段岸线变化特点、岸线变化与影响因素之间的关系。利用BP人工神经网络原理,提出了基于BP神经网络的河道岸线变形预测模型。在模型的建立过程中,对模型的影响因子进行了分析和确定,并对输入、输出因子的获取途径进行了探讨。提出了深泓线位置的估算方法,并进行了可行性验证。在预测过程中,探讨了模型训练样本数与隐含层节点数之间的关系,确定了训练样本数目。通过反复试验,确定了隐含层节点数、学习速率、动量系数等模型参数。同时,对人工建筑物对岸线变形预测的影响也进行了试验分析。预测结果表明模型适用于自然岸段的岸线变形预测,模型计算值与实际值吻合良好。从新的角度出发,把RS技术和ANN理论用于岸线演变分析,不但弥补了常规大范围测量的缺陷,而且解决了河岸线变形预测难的问题。它是对传统岸线演变方法的补充,也为信息化、数字化建设增加了新的内容。更多还原

【Abstract】 This paper is taken Zhuqi to Houguan of Min river as research subject, used bank-line information provided by the topographic map of surveying and RS image, statistics data of hydrology to analysis evolution of river bank-line. The main research works and achievements in the paper are as follows:Extract the RS information and surveying information on bank-line. Select suitable RS image to process recovering and strengthen on image. Use rational of bank-line RS and method of two value in order to extract the data and analysis the reason of error of RS bank-line. Through analysis the precision on RS bank-line to find out the error acquired on bank-line by RS technology is less. And the acquired data can be used in analysis of bank-line evolution.Analysis on the development of bank-line by using data of bank-line of surveying and RS bank-line information from Zhuqi to Honguan of Min river. The results reflect the features of bank-line from Zhuqi to Honguan of Min river and the relationship between the development of bank-line and influencing factors. With BP artificial neural network, a model is developed to predict river bank-line deformation based on BP neural network. During the process of setup the model, this paper analysis and determine the influencing factors, to discuss the acquiring path on input and output factors, to bring forward the predict method on position of thalweg, and to do the feasibility study. During the prediction, discussing the relationship between the sample size of model exercise and the number of node in hidden layers to determine the sample size of exercising. To determine the parameters of model on number of node in hidden layers, learning rate, coefficient of momentum. At the same time to test on the manual building influence on prediction of bank-line deformation. The predict result indicate the model is suitable for the prediction of bank-line deformation on natural bank-line, the calculated data is in consistent with the actual data.Start from the new point of view to put RS technology and ANN theory into analysis of bank-line development not only make up the shortcoming of normal big scale measurement, but also solve the problem of difficult to predict the deformation of bank-line. It is the supplement on traditional development of evolution on bank-line and bring new contents on the informative and digital construction.更多还原