基于Kullback-Leiber距离粗差探测的Bayes方法  

BAYESIAN APPROACH TO DETECTION OF GROSS ERRORS BASED ON DIVERGENCE OF KULLBACK-LEIBER

作　　者：王延停[1] 归庆明[1,2] 张倩倩[1] 

机构地区：[1]解放军信息工程大学理学院,郑州450001 [2]解放军信息工程大学测绘学院,郑州450052

出　　处：《大地测量与地球动力学》2012年第2期51-54,59,共5页Journal of Geodesy and Geodynamics

基　　金：国家自然科学基金(40974009);郑州市科技计划攻关项目(0910SGYG21198)

摘　　要：从影响分析角度,基于Kullback-Leiber距离提出了粗差探测的Bayes方法。对非等权独立观测情形,在一定先验分布下,对数据删除模型、方差膨胀模型和均值漂移模型进行了比较分析,给出3种模型相应的Kullback-Leiber距离的计算公式,建立了粗差探测的判别法则。最后将该方法应用于边角网的粗差探测,取得了较好的效果。With the view of influence analysis, a new Bayesian method for gross errors detection based on di- vergence of Kullback-Leiber is proposed. Under the condition of unequal weight and independent observations, a comparison among the data deletion model, the variance inflation model and the mean shift model based on certain prior distribution is made, and the computational formula of divergence of Kullback-Leiber about the three models is given and a judge rule of gross error detection is established. Finally, the method is used for gross error detection in triangulateration network and a good result is obtained.

关 键 词：BAYES方法 影响分析 Kullback—Leiber距离 粗差探测 方差膨胀模型 

分 类 号：P227[天文地球—大地测量学与测量工程] P207[天文地球—测绘科学与技术]