摘要
从一个新的角度出发,对应每个观测值引入一个识别变量,基于识别变量的后验概率提出一种新的粗差定位的Bayes方法,并构造相应的均值漂移模型给出粗差估算的Bayes方法。由于识别变量的后验分布往往是复杂的、非标准形式的,为此设计一种MCMC(Markov Chain Monte Carlo)抽样方法以计算识别变量的后验概率值。最后对一边角网进行了计算和分析。试验表明,本文给出的探测粗差的Bayes方法不仅充分利用了先验信息,而且克服了以往粗差定位方法的模糊性以及探测标准选择的问题,同时计算简便。
In terms of a new point, introducing a classification variable according to each observation, a new Bayesian method for positioning gross errors based on the posterior probability of the classification variable is put forward, and the mean shift model is constructed in order to get the Bayesian estimator for gross errors. Due to the posterior distribution of classification variable being complex and nonstandard, a MCMC（Markov Chain Monte Carlo）sampling method is designed to compute the posterior probability of classification variable. Finally, as an example, a side-angle adjustment network is computed and analyzed. Numerous experiments show that the method given not only makes use of prior information, but also overcomes the positioning vague ness of the existing detection methods and the difficulties of choosing the detection criterion. In the same time, this resolution is simple and quick with the result being perfect comparatively.
出处
《测绘学报》
EI
CSCD
北大核心
2008年第3期355-360,366,共7页
Acta Geodaetica et Cartographica Sinica
基金
国家自然科学基金项目(40474007)
国家杰出青年科学基金项目(40125013)
"基础地理信息与数字化技术"山东省重点开放实验室课题(SD040202)
河南省自然科学基金项目(0511010100)
关键词
BAYES方法
粗差
识别变量
后验概率
MCMC抽样
均值漂移模型
Bayesian method
gross errors, classification variable
posterior probability
MCMC sampling
mean shift model
