基于机器学习的多源矢量同名面实体几何不一致性识别和处理方法

A Method for Recognizing and Processing Geometric Inconsistencies in Multi-source Identical Polygonal Vector Entities Based on Machine Learning

针对多源矢量数据融合更新中同名实体之间存在明显几何位置差异,导致几何不一致性识别和处理自动化程度较低的问题,提出一种面向多源矢量同名面实体的几何不一致性识别与处理方法。首先,对几何不一致性的特征分类进行深入分析,通过机器学习方法构建基于同名面实体几何不一致性特征指标的识别模型;然后,引入点集配准算法进行实体位置对齐,进而实现同名面实体的几何一致性处理。试验选取舟山地区多源面状水系数据进行验证。结果表明,该方法几何不一致性识别结果具有较高的准确率,几何一致性处理结果优于现有直接叠置分析的方法,可以有效降低位置误差,提高几何一致性。

In the fusion and updating of multi-source vector data,there are significant geometric positional differences between identical entities,leading to a low degree of automation in geometric inconsistencies recognition and processing.A geometric inconsistency recognition and processing method for multi-source identical polygonal vector entities is proposed in this paper.Firstly,an in-depth analysis is conducted on the classification of geometric inconsistency features,and a recognition model based on geometric inconsistency feature indices of identical polygonal entities is constructed by using machine learning methods.Then,the point set registration algorithm is introduced for entity alignment to achieve geometric consistency processing of identical polygonal entities.The effectiveness of the proposed method for multi-source hydrological data are tested in Zhoushan area.The results demonstrate that this method has high accuracy in identifying geometric inconsistencies,and the geometric consistency processing results are superior to those of existing direct overlay operation methods.Additionally,this method can effectively reduce positional errors and improve geometric consistency.