摘要
针对常用直线拟合方法在数据存在粗差或异常值扰动时,存在拟合结果不稳定的缺点,提出一种稳健的直线拟合方法。该法以整体最小二乘法为基础,在考虑全部观测量存在误差的情况下,通过利用一定的准则删除数据中的粗差或异常值,从而获得稳健的直线参数。实验结果表明,稳健整体最小二乘直线拟合不仅考虑了全部观测值中的误差,而且能剔除数据中的粗差或异常值,精度更高,结果更为可靠。
In traditional line fitting methods, the gross error and outliers are not considered, and thus the results of parameters estimation are not accurate. In order to overcome this shortcoming, a robust method for line fitting was proposed. The method is based on total least squares (TLS). In consideration of errors in all observations, by deleting outliers of point clouds according to some criteria, a robust solution to line fitting parameters is obtained. Analytical simulation experiment was conducted, and comparison between the traditional method and TLS method was also implemented. The results show that the method has the capability to overcome bad influence from outliers, and to increase the reliability of parameter estimation.
出处
《工程勘察》
CSCD
2012年第2期60-62,共3页
Geotechnical Investigation & Surveying
基金
国家自然科学基金资助(40874010
40940011)
江西省自然科学基金(2008GQC0001
2010GZC0009
2010GZC0008)
关键词
直线拟合
整体最小二乘
粗差
异常值
稳健
line fitting
total least squares
gross error
outliers
robustness