基于WTLS的部分最小二乘粗差探测法

Partly Least Square Outlier Detection Method Based on WTLS

针对平差时EIV模型中所含粗差的问题,提出了EIV模型的粗差探测法,利用变形后的加权总体最小二乘法依次对假设不含粗差的一组观测向量进行平差计算,然后通过平差后所得到的单位权方差估计值的大小对观测向量进行分组,从而达到粗差定位的目的。通过算例分析比较,得出在观测值独立等精度时,本文的方法可以有效的发现观测数据所存在的粗差。

For the problem of outliers in EIV models, a new outlier detection method of EIV model is introduced. Assuming a set of observation vector without outliers is adjusted orderly,then using the weighted total least squares after deformation. To detect outliers in observations, the estimation of unit weight variance can be used to classify observations. The numerical example reveals that the new method can effectively locate the outliers in observations when the observations have the same weight.