基于BP神经网络和拓扑参数的道路网选取研究

Selection of Road Network Using BP Neural Network and Topological Parameters

道路网选取是自动制图综合的重点和难点之一,运用智能化方法实现选取是当前研究的热点。BP神经网络具有强大的非线性映射能力,可以模仿人脑机能,通过对样本的学习和训练实现自动选取;结合拓扑参数,可以使选取结果很好地保持原道路网的连通性和整体结构特征。因此,提出一种基于BP神经网络和拓扑参数的道路网选取方法。首先选择训练样本并计算其拓扑参数;然后设计BP神经网络的结构,利用训练样本进行训练,找出最佳网络结构;最后选取不同特征的道路网进行实验,将选取结果与专家选取的结果进行对比分析,评价了该方法的优势与不足,并指出了下一步的改进方向。

The selection of road network is one of the key and difficult tasks in automatic map generalization, and the selection using intelligent methods is very popular in current research.The BP neural network has strong ability of nonlinear mapping thatcan imitate the functions of human brains and realize the automatic selection by machine learning. It can also keep the connections and the whole structure of original road networks by using topological parameters. So a method of road network selection using the BP neural network and topological parameters was proposed. Firstly, we selected the training samples and calculatedtheir topological parameters; Secondly, we designed the structure of the BP neural network and trained it using the training samples to find the optimal network structure; Finally, an experiment with various type of road network was given. Based on the comparison between the results of experiment and the results from experts, the advantages and shortcomings were analyzed, and the future work to improve this approach was discussed.