径向基函数神经网络的路网自动匹配算法

针对VGI数据中检测更新的问题,该文提出基于径向基函数的神经网络自动匹配算法。通过选取路段的距离、方向、形状和长度4个空间特征的相似度作为衡量路段是否匹配的指标。考虑到4个空间特征指标对匹配的影响力不同,在RBF(radial basis function)神经网络中的隐含层对基函数引入粒度拉伸因子,使径向对称的RBF顾及各向异性。同时对输出层在线性加权求和函数的基础上引入sigmoid函数,使计算结果(路段的匹配度值)归一化。该算法对数据质量较差的VGI路网具有很好的匹配能力,与BP神经网络相比,RBF神经网络在地图匹配中具有更好的匹配效率。

Automated matching road networks utilized RBF neural network

Map matching is an important prerequisite for map integration,map updating,change detection and so on.The amount of volunteered geographic information (VGI)has increased in recent years,and VGI is drawing great attention with its superiority for map updating.In order to achieve the matching of VGI road network and the professional road network,this paper proposes a map matching algorithm based on radial basis function (RBF)neural network.It takes distance,orientation,shape and length as the evaluation factors to determinate whether the two road arcs are the same road or not.Add a scaling parameter to the basis function of the hidden layer to make the RBF neural network anisotropy and utilized sigmoid function in the output layer to make the calculation results normalization.Firstly,select samples from the professional road data and the VGI road data,calculate the four feature similarities of these samples,these four similarities with addition of the matching rate serving as learning modes.Secondly,calculate the mean value and variance of each feature similarities.Thirdly,input the learning modes to train RBF neural network,get the connection weights between the hidden layer and the output layer.At last,input the VGI road network and the professional road network,calculate the matching rate of road arcs.Result shows that our algorithm achieves good performance even the quality of VGI data is not so good,besides,utilized RBF neural networks has a better efficient than utilized BP neural network in matching road networks.