Real-time SWIR hyperspectral imaging with polarimetric capability

Abstract

Traditional hyperspectral imaging (HSI) sensors are inherently time-sequential during capture, relying upon scanning techniques to construct the resultant hypercube. This temporal constraint hence restricts the use of HSI to static scenes or platforms. The novel sensor outlined within this paper enables snapshot HSI. The Near-Infrared Image Replicating Imaging Spectrometer (N-IRIS) operates without any rejection in polarized light. This prototype has eight SWIR bands and a diagonal FoV of two degrees, with potential for sixteen bands in other infrared regions. Unlike other snapshot techniques, N-IRIS produces a spectral image directly without inversion. Many additional benefits include inherent compactness, robustness, no-moving-parts operation, lower processing overheads and resource needs. Dual polarimetricspectral imaging is also possible due to its inherent design, which offers additional discrimination and higher throughput. HSI algorithms for anomaly detection are prolific in variety, but almost none of them consider the temporal dimension, mainly due to current limitations on speed. This paper describes the results from advanced algorithms implemented on COTS hardware for video-rate operation and designed to exploit the temporal dimension. The synergy with N-IRIS has achieved anomaly detection within streaming HSI hypercubes at video frame-rates. Recorded datasets include static ground scenes with transient targets, while further AVRIS imagery achieved the video-rate detection of embedded simulated targets therein. This new capability through N-IRIS hence broadens the potential application and benefit of HSI sensors to dynamic or transient situations.