中文摘要：

介绍了数据预处理对基于径向基函数网络的高程转换的必要性，提出了基于径向基函数网络高程转换建模的基本原理和实现步骤。结合工程实例，重点对归一化、中心化、标准化3种不同的数据预处理方法对基于径向基函数网络高程转换的影响进行计算分析。结果表明：针对大面积、高程异常值变化较大的区域，经过数据预处理后模型能够达到更高的转换精度；不同的预处理方法对转换结果影响不同，中心化预处理较其他两种方法能够更大程度地提高径向基函数网络高程转换的精度；归一化预处理方法不一定是最优的，进行高程转换前有必要对预处理方法进行选择。

英文摘要：

The necessity of data pretreatment to the elevation fitting on radial basis function network modeling is described.The basic principle and implementation steps of the conversion modeling is depicted based on radial basis function networks elevation.With engineering examples,normalization,centralization,standardization of three different data pretreatment methods is used to calculate the impact of good conversion on radial basis func-tion network.It is shown that data pretreatment model can achieve better conversion accuracy in a large area with elevation changes over the outliers.Different pretreatment methods result in different conversion.The cen-tralization can improve the elevation accuracy on radial conversion function network in a greater degree.Normal-ization wouldn＆#39;t be the best data pretreatment method.It is necessary for the selection of pretreatment method before elevation conversion.