基于MBR组合优化算法的多尺度面实体匹配方法

针对多尺度匹配中同名实体位置偏差较大，无法直接通过面积重叠法获得候选匹配的问题，本文提出了一种基于最小外包矩形（MBR）组合优化算法的多尺度面实体匹配方法。本文方法的基本思想是通过MBR组合优化和简要的形状特征来筛选1∶1、1∶N和M∶N候选匹配，然后构建多因子人工神经网络模型来评估候选匹配。试验选取浙江省舟山市1∶2000岛礁基础数据和1∶10 000陆地基础数据中的居民地与设施面进行匹配算法的验证。结果表明，本文方法相对于基于面积重叠-神经网络的匹配方法表现出显著的优势，对存在位置偏移的匹配数据准确率和召回率分别达到了达到96.5%，达到89.0%，且能够识别所有匹配类型。

A Multi-scale Polygonal Object Matching Method Based on MBR Combinatorial Optimization Algorithm

Aiming to solving the problem of positional discrepancy of corresponding objects in multi-scale polygonal object matching and that the potential matching pairs can't be directly identified by the method of areal overlapping,it is proposed that a multi-scale polygonal object matching method based on minimum bounding rectangle combinatorial optimization algorithm .The basic idea of our method is that:①identifying the potential matching pairs of 1:1,1:N andM:N with combinatorial algorithm and simple shape characteristic;②establishing multi-characteristic artificial neural network model to evaluate these potential matching pairs.The proposed method is demonstrated in the experiment of matching between 1:2000 and 1:10000 polygonal objects of residential buildings and industrial facilities in Zhoushan, Zhejiang Province.The experimental results showed that the proposed matching method show superior performance against a method of area overlapping and artificial neural network.Its precision and recall are 96.5％ and 89.0％ under the positional discrepancy scenario,and it successfully match 1:0,1:1,1:N andM:N matching pair.