摘要
在鲁棒H∞滤波应用过程中,如果量测序列含有野值,将会严重影响滤波精度。针对这一问题提出一种新的剔除野植的方法;从新息入手,首先利用小波变换系数特性,通过最细尺度上的小波系数来检测野值点,然后基于信息扩散原理,采用替代方法,对含有单个或连续野值的新息加以修正,从而达到检测和剔除野值的目的。通过对基于MEMS的车载微惯性SINS/GPS组合导航的仿真表明,新算法能够有效的检测出野值,并在野值单个或成片出现的情况下都能保证滤波精度。
In the light of practical application the outliers makes the estimation of filtering inaccurate. A new outliers restraining method is proposed. Based on the characteristic of new information, first the coefficient characteristic of wavelet transform is used. The wavelet coefficient of the fine scale is used to detect outliers, and the method based on information pervasion theorem is used to correct the single outlier or series outliers. The simulation of MEMS- SINS/GPS integrated navigation in the long-distance vehicle system shows that the new algorithms can detective outliers effectively and ensure the precision of filtering exactly.
出处
《传感技术学报》
CAS
CSCD
北大核心
2012年第6期859-863,共5页
Chinese Journal of Sensors and Actuators
基金
国家自然科学基金项目(50805004)
国家973计划项目(2011CB711106)
关键词
野值检测
野值剔除
小波变换系数
信息扩散
新息修正
outliers detecting
outliers eliminating
wavelet transform coefficients
information pervasion
innovation revising
