中文摘要：

常规确定区域似大地水准面GPS水准法最常用的曲面拟合法通常只考虑测量误差，导致拟合结果不准确，并且简单地利用内、外符合精度去评定一个模型的拟合精度不是很可靠，只能在一定程度上说明模型拟合效果。针对此问题，利用同时顾及测量误差和模型误差的最小二乘配置法进行高程拟合，将其拟合得到的精度值作为参考，并提出利用高程异常残差三维图和内、外符合精度同时进行精度分析，然后与二次多项式和BP神经网络拟合结果进行对比分析。研究结果表明：对于大面积高程异常值变化较大的地区，最小二乘配置法能够取得较高的拟合精度；利用残差值三维图进行精度评定是一种有效可靠的方法。

英文摘要：

The common method to determine Quasi-Geoids is GPS leveling.The fitting method is surface fit-ting,considering only the measurement error,leading to inaccurate fitting result.It is impossible to judge fitting result by inner and outer fitting,as it can only explain a certain extent of the fitting effort to the model.Thus the least square collocation method is used for elevation fitting which takes measurement error and model error into account.By taking the fitting result as a reference,and method of using three dimension diagrams of residual of elevation anomaly,and the inner and outer fitting result is put forward to analyze the results.Then the result with quadratic polynomial and BP neural network is compared.The result shows that LSC method can achieve high fitting precision for large area with height anomaly value changes significantly.It is an efficient approach to assess the fitting result with the three dimensional diagram.