Range invariant anomaly detection for LWIR polarimetric imagery

Abstract

In this paper we present a modified version of a previously proposed anomaly detector for polarimetric imagery. This modified version is a more adaptive, range invariant anomaly detector based on the covariance difference test, the M-Box. The paper demonstrates the underlying issue of range to target dependency of the previous algorithm and offers a solution that is very easily implemented with the M-Box covariance test. Results are shown where the new algorithm is capable of identifying manmade objects as anomalies in both close and long range scenarios.