Title :
Statistics for characterizing data on the periphery
Author :
Theiler, James ; Hush, Don
Author_Institution :
Space & Remote Sensing Sci., Los Alamos Nat. Lab., Los Alamos, NM, USA
Abstract :
We introduce a class of statistics for characterizing the periphery of a distribution, and show that these statistics are particularly valuable for problems in target detection. Because so many detection algorithms are rooted in Gaussian statistics, we concentrate on ellipsoidal models of high-dimensional data distributions (that is to say: covariance matrices), but we recommend several alternatives to the sample covariance matrix that more efficiently model the periphery of a distribution, and can more effectively detect anomalous data samples.
Keywords :
Gaussian processes; data handling; object detection; security of data; Gaussian statistics; anomaly detection; data distributions; distribution periphery; ellipsoidal models; target detection; Covariance matrix; Data models; Ellipsoids; Hyperspectral imaging; Robustness; Support vector machines; Gaussian mixture models; anomaly detection; expectation-maximization; leptokurtosis; outlier; probability distribution; robust statistics; target detection;
Conference_Titel :
Geoscience and Remote Sensing Symposium (IGARSS), 2010 IEEE International
Conference_Location :
Honolulu, HI
Print_ISBN :
978-1-4244-9565-8
Electronic_ISBN :
2153-6996
DOI :
10.1109/IGARSS.2010.5651361
