Segmentation Adaptive RX: an Algorithm for SpectralAnomaly Detection in a Variety of Measured-RadianceConditions

摘要

Hyperspectral anomaly detection algorithms are well developed however their ability to account forillumination conditions is limited only to mild variations. We propose an approach specificallydesigned to handle shadows and poorly illuminated regions present in otherwise well-illuminatedimagery without making any assumptions about shaded backgrounds or object signature evolution.The algorithm, Segmentation Adaptive RX (SARX), relies on panchromatic segmentation of the datainto dark and bright clusters based on the illumination level. Bright cluster detection employs standardsubspace RX and dark cluster detection subspace is limited by only few higher variance spectraldimensions to reflect diminished signal-to-noise ratio in shadows. Anomaly detection near thegeographical border between the clusters utilizes Stochastic Mixing Model. We demonstrateexperimentally superior ability of SARX to detect anomalous objects in variety of illuminationconditions.