论文部分内容阅读
在综合考虑多个特征因子的线要素匹配时,根据经验知识确定各特征因子的权值会造成人为误差。针对该问题,本文提出了基于人工神经网络的多特征因子路网匹配算法,根据线要素的几何和拓扑特性选取长度、方向、形状、距离及拓扑5个特征因子的相似度作为路网匹配参考因子。首先,分别在参考图层和待匹配图层中选取样本数据组成样本对,计算样本数据的5个特征因子相似度,用样本数据的5个特征因子相似度和样本的匹配度组成学习模式对;然后,利用BP神经网络的误差反向传播机制自动学习调整各神经层之间的连接权值;最后,输入全部数据,计算参考图层的弧段和待匹配图层的弧段间的匹配度,实现综合多特征因子的路网匹配。实验结果表明,利用人工神经网络进行综合多特征因子的路网匹配可以提高匹配效率和匹配准确度。
When comprehensively considering the line feature matching of multiple feature factors, it is human error to determine the weight of each feature factor based on empirical knowledge. In order to solve this problem, this paper proposes a multi-factor factor road network matching algorithm based on artificial neural network. According to the geometric and topological characteristics of line elements, the similarity of five feature factors of length, direction, shape, distance and topology is selected as the road network matching reference factor. First of all, we select the sample pairs from the reference layer and the layer to be matched, calculate the similarity of the five feature factors of the sample data, construct the learning pattern pair by using the five feature factor similarity of the sample data and the sample matching degree Then, the error backpropagation mechanism of BP neural network is used to automatically adjust and adjust the connection weights between all neural layers. Finally, all the data are input to calculate the matching between the arcs of the reference layer and the arcs of the layers to be matched Degree, to achieve a comprehensive multi-factor road network matching. The experimental results show that the matching of road network with artificial neural network can improve the matching efficiency and matching accuracy.