Variable subspace model for hyperspectral anomaly detection

Detection of anomalies in a large area is an important objective in remote sensing using hyperspectral imaging system. Some conventional anomaly detectors for such an application have a user-specified parameter which is the dimension of the clutter subspace. The range of possible values for this parameter is typically large. In this paper, an anomaly detector with a different parameter is proposed. The anomaly detector partials out the effect of the clutter subspace from a pixel by predicting each spectral component of a pixel using a linear combination of the clutter subspace. The dimension of the clutter subspace can vary from one spectral component to another one and is determined by a parameter based on the maximized squared correlations. The Mahalanobis distance of the resulting residual is defined as the anomaly detector. The experimental results are obtained by implementing the anomaly detector as a global anomaly detector in unsupervised mode with background statistics computed from hyperspectral data cubes with wavelengths in the visible and near-infrared range. The results show that the parameter in the anomaly detector has a significantly reduced number of possible values and the values can be determined automatically in an iterative mode.