神经网络决策树的矢量数据变化信息快速识别方法

Research on Change Information Recognition Method of Vector Data Based on Neural Network Decision Tree

针对GIS数据更新中变化信息的自动提取问题,提出基于神经网络决策树的变化信息快速识别方法。设计了基于四叉树的变化信息层次检测算法,通过比较对象的节点-弧段特征快速定位到变化区域。以新旧要素的匹配特征为依据,通过神经网络决策树实现变化信息的识别,兼顾了决策树实现效率高和神经网络的自适应处理的特征。1∶2000地形图变化信息识别试验结果表明:该方法计算效率高,能够准确快速地识别出变化信息,有助于提高GIS数据库的动态更新的自动化与智能化水平。

To realize automatic extraction of change information in GIS vector data updating, a change information recognition method based on neural network decision tree is proposed. A change information hierarchical detection method based on quad-tree was designed and realized in this paper. With the comparison of vertex-edge characteristics, this method could rapidly locate change areas. Based on the matching of correspondent objects, a neural network decision tree method was applied to recognize the change information. This method combined the effective logical judgment of decision tree and adaptive processing of neural network. 1 ： 2000 Topographical data were used to verify the effectiveness of the method. Experimental result shows that this method can achieve high computing speed and effectively detect the change pattern of vector data, which can improve the automation and intelligence level of dynamic updating in GIS database.