基于LWR-Pettitt算法的GNSS变形信息的识别与预警

针对GNSS变形监测的数据通常无特定的分布特性,不利于建立具有普遍性的统计量,本文引入非参数检验中的Pettitt检验方法建立统计量,进行GNSS监测序列中变形信息的识别与预警。同时,针对Pettitt检验方法中存在统计量振幅波动较大,并且只能识别单个变形信息的问题,基于局部加权回归法LWR,对Pettitt检验中统计量与阈值判断方法进行优化,提出一种新的LWR-Pettitt算法,并用于GNSS变形信息识别与预警。试验结果表明,对于不同的测试数据,新方法均可有效地识别变形信息的发生位置,特别对于趋势型变形;而对于突变型变形,新方法可有效地识别2倍标准差以上的连续变形信息。

Recognition and early warning of deformation information in GNSS coordinate series based on the LWR-Pettitt method

GNSS deformation monitoring data usually has no specific distribution characteristics, which is not conducive to the establishment of universal statistics. As a non-parametric test method, the Pettitt test is introduced to establish statistics of GNSS deformation monitoring data, and then the deformation information identification and early warning are carried out. However, the statistics constructed by the Pettitt test always have large fluctuation, and can only be used to identify the single deformation information. Therefore, a new method named LWR (Locally Weighted Regression)-Pettitt is proposed to optimize the statistics and threshold judgment method in the Pettitt test, and is used for GNSS deformation information identification and early warning in this paper. The experimental results show that the proposed method can identify the location of deformation information effectively, especially for trend-type deformation information, but for the abrupt-type deformation information, it can only identify the continuous deformation information more than two-fold standard deviation equivalent deformation.