GPS坐标时间序列粗差剔除方法比较分析

Comparative analysis on outlier elimination methods for GPS coordinate time series

为了更加有效地探测和剔除GPS台站位移序列的粗差,对几种常见的GPS坐标时间序列粗差剔除方法进行比较分析:分别采用GPS坐标时间序列中常用的粗差探测方法如5倍中误差法(5σ)、3倍中误差法(3σ)、中位数绝对偏差法、四分位距法对GPS坐标时间序列进行粗差处理分析;并以40个GPS基准站坐标序列为对象进行粗差剔除比较;最后对去除粗差后的GPS坐标序列进行噪声模型分析,总结粗差剔除前后坐标时间序列的最佳噪声模型。实验结果表明,4种粗差探测和剔除方法都能检测出时间序列中的粗差,且以四分位距法的效果最为明显。

In order to further detect and eliminate the gross error of global positioning system(GPS)station displacement sequences,the paper comparatively analyzed several common outlier elimination methods of GPS coordinate time series:the common outlier detection methods in GPS coordinate time series,such as 5 times mean square error method,3 times mean square error method,median absolute deviation method and interquartile range method,were used to process and analyze the outlier;and 40 GPS reference station coordinate series were taken as the objects to compare the eliminated outliers;then the noise models of GPS coordinate time series after outlier removal were analyzed;finally the optimal noise model of coordinate time series before and after outlier elimination was summarized.Experimental result showed that all four outlier detection and elimination methods could detect outlier in the time series,among which interquartile range method would be most effective.