欢迎您！东篱公司
退出
-
申报数据库
-
立项数据库
-
成果数据库
-
项目获奖数据库
中文摘要：
居民地匹配是多源空间数据融合和多尺度数据更新的核心技术环节。针对居民地匹配算法中出现的指标权重、匹配判定的总相似性阈值和各指标相似性阈值的准确量化难题,引入人工神经网络技术,利用人工神经网络在处理多要素、复杂性、模糊性分类问题上的优势,将形状相似度、方向相似度、位置相似度、大小相似度和重叠面积相似度作为输入,采用人机结合的神经网络训练策略,对3层BP神经网络进行训练,针对不同的匹配场景获取神经网络的权重向量集,实现了多指标综合衡量的居民地匹配。实验表明,该方法解决了多指标匹配算法存在的理论严谨性问题,回避了权重和阈值准确设置的难题,保证了匹配算法的科学性、稳定性和准确性。
英文摘要：
Settlement matching is one of the kernel parts of multi-source spatial data fusion and multi-scale data updating.But it is hard to accurately quantitate the weights and thresholds of every feature which involves in the matching algorithm.So the ANN（Artificial Neural Network） technique was introduced for its advantages in solving problems with complexity,illegibility and multi-features characters.The similarity of shape,orientation,position,size and overlap area were imported to three layers BP neural network.And the weight matrix was achieved according with every matching scene after training it with human computer cooperation method.Test illustrated that this method effectively improved the matching precision and stability with successfully overcoming the shortage of original algorithms which need weight and threshold setting.
同期刊论文项目
同项目期刊论文
期刊信息
位置：