观测数据野值剔除的TBS-LS逼近方法

摘要

This paper presents an outlier identification and elimination algorithm with least-square triangle B-Spline based on the theory of approximation. Triangle B-Spline with both whole and local properties of superiority is utilized and constructional algorithm analysis of error residual and skewness is given. Triangle B-Spline curve fitting with the given norm in the evaluation system enables better identification and elimination of outliers in observation data. Finally, the effectiveness of the method is illustrated with some examples.%提出了以最小二乘逼近方法为基础的数据野值判别与剔除算法.利用TBS(三角B样条)曲线同时具有局部性和整体性的优越性构造最小二乘拟合算法,并结合偏度分析与残量分析误差方法,在给定范数意义下的评价系统中,可以得到TBS-LS(最小二乘三角B样条)拟合曲线,从而可以更好地识别并剔除野值.最后给出算法以及主要结果,通过实例说明方法的有效性.