Evaluation of the sub-pixel performance of anomaly detectors

Anomaly detection in hyperspectral data has received much attention for various applications and is especially important for defense and security applications. Anomaly detection detects pixels in the hyperspectral data cube whose spectra differ significantly from the background spectra. Most existing methods estimate the spectra of the (local or global) background and then detect anomalies as pixels with a large spectral distance w.r.t. the determined background spectra. Many types of anomaly detectors have been proposed in literature. The most well-known anomaly detector is the RX detector that calculates the Mahalanobis distance between the pixel under test and the background. This paper investigates the sub-pixel detection performance of two classes of anomaly detectors: the family of RX-based detectors and the segmentation-based anomaly detectors. Representative examples of each class are selected and results obtained on three different datacubes are analyzed.