Outlier Detection in Hyperspectral Imagery Using Closest Distance to Center with Ellipsoidal Multivariate Trimming
Versandkostenfrei!
Versandfertig in über 4 Wochen
53,99 €
inkl. MwSt.
PAYBACK Punkte
27 °P sammeln!
Many multivariate techniques are available to find outliers in a hyperspectral image. Among the algorithms one may utilize is a global anomaly detector called Ellipsoidal Multivariate Trimming (MVT). In this paper we tested the efficacy of using the Closest Distance to Center (CDC) algorithm in conjunction with MVT to find outliers among a hyperspectral image. Since MVT is a global anomaly detector the images were first clustered using a variety of techniques. Among the hyperspectral images used for evaluation in this study, only one of the images contained more than 5% outliers in any given c...
Many multivariate techniques are available to find outliers in a hyperspectral image. Among the algorithms one may utilize is a global anomaly detector called Ellipsoidal Multivariate Trimming (MVT). In this paper we tested the efficacy of using the Closest Distance to Center (CDC) algorithm in conjunction with MVT to find outliers among a hyperspectral image. Since MVT is a global anomaly detector the images were first clustered using a variety of techniques. Among the hyperspectral images used for evaluation in this study, only one of the images contained more than 5% outliers in any given cluster set. Based upon the assumption that this is normally the case for most images, the standard use of 50% retention within MVT does not perform as well as using a higher value such as 95% for retention in MVT. This use of a higher number of observations for the estimate of the mean and covariance is shown to decrease the effect of swamping seen when using 50% retention. Furthermore, the use of CDC to initialize the MVT iteration process did not have any effect on outlier determination, but did increase the time to compute significantly.
Andere Kunden interessierten sich für
