CAJ | 학술논문

针对水下无人潜航器（ UUV）预测跟踪过程中传感器所采集数据的不准确问题，提出了一种利用小波隐马尔可夫模型进行UUV预测跟踪数据孤立点检测的方法。利用改进递归小波变换，对原始数据进行重构，小波系数中孤立点处的系数得到局部放大。小波系数的更新基于历史时刻的数据，因此，可以利用正常数据与孤立点的差异对数据进行实时处理。利用隐马尔可夫模型定义异常值分布判定函数，并以其作为依据，检测特征明显异于正常数据的孤立点。将准孤立点利用惰性算法进行重检测，提高孤立点检测的准确性。湖试数据验证了该方法能够有效地检测出UUV预测跟踪中的数据孤立点。
침대수하무인잠항기（ UUV）예측근종과정중전감기소채집수거적불준학문제，제출료일충이용소파은마이가부모형진행UUV예측근종수거고립점검측적방법。이용개진체귀소파변환，대원시수거진행중구，소파계수중고립점처적계수득도국부방대。소파계수적경신기우역사시각적수거，인차，가이이용정상수거여고립점적차이대수거진행실시처리。이용은마이가부모형정의이상치분포판정함수，병이기작위의거，검측특정명현이우정상수거적고립점。장준고립점이용타성산법진행중검측，제고고립점검측적준학성。호시수거험증료해방법능구유효지검측출UUV예측근종중적수거고립점。
A method for outlier detection based on wavelet HMM (hidden Markov model) is proposed in this paperin order to deal with the inaccurate original data collected from sensors during UUV forecast tracing .The improvedrecursive wavelet transform (IRWT) is used to reconstruct the original data and locally amplify the wavelet coeffi -cients of the outliers.The update of wavelet coefficients are based on the data at the historical moment , thus, datamay be processed in real time by utilizing the difference between normal data and outliers .The judgment function onthe distribution of abnormal values is defined by HMM .In addition, on the basis of this, the outliers with featuresobviously different from the normal data are detected.The quasi-outliers are redetected by using a lazy algorithm forimproving the accuracy of the detection results.The data from the lake experiment verify that the method may effectivelydetect the data outliers in UUV forecast tracing.