基于RBF神经网络的OSM道路网智能选取

Intelligent Selection of OSM Road Network Based on RBF Neural Network

针对目前自动制图综合方法在软件平台没有普及,电子地图仍以人工编制为主,Open Street Map(简称OSM)道路不适合作为标准地图等问题,提出了一种将径向基函数神经网络(Radial Basis Function, RBF)应用在OSM道路网自动选取中的方法。根据多种常用的语义、几何、拓扑参数综合考虑道路网的重要性,并将该算法在系统平台中加以实现,以期为OSM道路网自动选取的精度提高和应用普及提供解决方案。实验结果与实际制图结果在形状结构上保持良好,精度为86.92%,相较于BP(Back Propagation)神经网络算法的效果有所提高。

Open Street Map (OSM) data is not suitable to be used as a standard map. Considering this, in addition to the situation that automated cartographic generalization methods are not popular in the software platform and the compilation of electronic map is still mainly manually, a radial basis function neural network (RBF) is built to apply in the automatic selection of the OSM road network. Combining self-organization and supervised learning, RBF learns faster than the commonly used back-propagation neural network. Taking into account the importance of the commonly used semantic, geometric, topological parameters of the road networks, an algorithm is developed and then implements in the software platform in order to achieve the solution with high precision selection of the OSM road network. The experiments with OSM road network show that the experimental results and the standard map are in good shape with the accuracy of 86.92%, which is better than BP algorithm.