Hyperspectral imagery has traditionally been limited to a relatively narrow wavelength range of approximately one octave. Thus constrained, remote-sensing research involving hyperspectral images fails to capture the full complement of wave phenomenology. While observing transitions in the mid-wavelength IR (MWIR) spectrum, for example, important molecular transitions may be missed in long-wave IR, and vice versa. Similarly, one might imagine ‘ machine vision’ techniques that rely on acquiring spectra in the LWIR that correspond to peaked thermal emissions from cooler areas on a manufactured item or, with respect to the MWIR, from hot spots that signify a failure.
In the past, researchers have invoked two or more focal plane arrays (FPAs), each with an ‘optical feed’ from a grating or prism optimized to provide higher efficiency hyperspectral coverage over restricted ranges (see Figure 1). A complicating factor here is that using one or more beam splitters to segregate the imaged scene into the various wavebands is a complex undertaking, particularly at cryogenic temperatures. One creative team of astronomers used a combination of a grating and prism, with dispersion axes oriented perpendicularly to segregate the multiple grating orders into different spatial locations on a single FPA. This ‘cross-dispersion spectrometer’ enjoyed a measure of success, but the restriction it imposes on the spatial dimension of hyperspectral images is considerable.
Figure 1. Traditional, dual-channel spectrometer.
We have recently taken a different approach that uses the various grating orders but lets the multiwaveband focal plane register and keep them separate. This has enabled perfect registration with considerable reductions in mass, volume, and the cryocooling requirement.
Dual-band FPAs were originally developed for multispectral imaging applications, for which their compactness and band-to-band pixel registration relative to conventional devices were clear advantages. As the architecture matured for quantum-well and mercury cadmium telluride FPAs, and has now become plausible for strained layer superlattice technology, additional potential uses include hyperspectral imaging.
For the applications that employ gratings, the different orders can sometimes be paired with wavebands of the dual- (or multi-) FPA, enabling high-efficiency imaging over broad regions. One such pairing of three-band hyperspectral performance with regions of high atmospheric transmission is shown in Figure 2. Exploiting this ‘third dimension’ of FPA imaging layers for dual- and multiwaveband capability has proved useful for multiwaveband imaging. In addition, we have now demonstrated similar advantages for hyperspectral applications.1
Figure 2. An example of how three grating orders can usefully fit into available atmospheric transmission bands.
Our strategy thus involves recording two grating orders by allowing the MWIR spectrum to be captured by the MWIR layer of the FPA, and similarly for the LWIR with its grating order and FPA layer. The case for this procedure becomes more compelling insofar as a grating blazed for high efficiency at a given wavelength also achieves peak efficiency in the next higher grating order at one-half the first-order wavelength. These peak wavelengths are dispersed by the grating at the same angle.
It is also highly advantageous for some applications to have ‘perfectly registered’ dual-band spectra as a result of overlapping grating orders sampled by dual-band pixels in perfect octaves of wavelength. This leads to interpretation of the spectral images without requiring extensive re-gridding of pixels to properly register wavelengths for a given spatial sample, a considerable simplification
Another promising hyperspectral approach is based on the computed tomographic imaging spectrometer (CTIS), which has been extensively investigated for single-waveband FPAs at the University of Arizona and also at the Photonics Research Center of the United States Military Academy. More recently, researchers have pursued ‘dual-band’ CTIS concepts in the IR,2 noting that the computer-generated holographic phase grating traditionally used as a transmissive dispersion device for the CTIS also provides additional spectral orders that can be exploited with dual-band IR FPAs. The challenge now is to design a dual-band CTIS that makes maximal use of similar existing FPA capability in terms of pixel format and sensitivity.
Space Vehicles Directorate,
Air Force Research Laboratory (AFRL)
Kirtland AFB, NM
Paul D. LeVan is a senior physicist and technical advisor, and past head of AFRL's space focal plane array group. His achievements include overseeing the export of IR detectors to the European Space Agency's space astronomical telescope, participating in the definition and procurement of IR sensors for the Air Force Advanced Electro Optical System, and sponsoring the development of dual-band IR FPAs.