Kalman滤波几种异常检验法  

Methods for Outlier Detection of Kalman Filtering

作　　者：聂建亮[1] 吴富梅[2] 

机构地区：[1]长安大学地质工程与测绘学院,陕西西安710054 [2]信息工程大学测绘学院,河南郑州450052

出　　处：《地球科学与环境学报》2008年第2期204-208,共5页Journal of Earth Sciences and Environment

基　　金：国家自然科学基金项目(4027400240474001);地球空间环境与大地测量教育部重点实验室项目(1469990324233-04-11)

摘　　要：为了检测动态导航观测异常和动力学模型异常,采用预测残差构造观测误差和动力学模型误差整体检验法,对观测异常可分别采用以模型为准的观测异常检验、以当前历元可靠观测为基准的异常检验或以Kalman滤波估值为基础的异常检验的方法;对于动力学模型异常检验,可以分别采用状态不符值检验法、以状态参数Kalman滤波估值为基础的动力学模型异常检验或以可靠观测为基础的动力学模型误差整体检验的方法。分析了以上几种检验方法的特点,并用实测数据进行了检验。结果表明:在观测异常或动力学模型异常处,异常检验对导航数据精度有一定程度地提高。In order to detect the outliers of measurements and dynamic models, the predicted residual vector is used to construct methods to the integral error detection of measured outliers and dynamic model error. There are three types of statistics, that is, the outlier detection based on the credible observation, the reliable measurements at present epoch, and the estimations of Kalman filtering respectively. For the discrepancies of states, the estimations of Kalman filtering and the reliable measurements are used to detect dynamic model error. These methods are compared and analyzed with the data of navigation. It is proved that the outlier detection can improve the precision when measurement outliers or dynamic model errors exist.

关 键 词：观测异常 动力模型异常 异常检验 KALMAN滤波 

分 类 号：P207.1[天文地球—测绘科学与技术]