A fast and precise registration method for multi-image snapshot Fourier transform imaging spectroscopy is proposed. This method accomplishes registration of an image array using the positional relationship between homologous points in the subimages, which are obtained offline by preregistration. Through the preregistration process, the registration problem is converted to the problem of using a registration matrix to interpolate subimages. Therefore, the hardware interpolation of graphics processing unit (GPU) texture memory, which has speed advantages for its parallel computing, can be used to significantly enhance computational efficiency. Compared to a central processing unit, GPU performance showed ∼27 times acceleration in registration efficiency.

A prototype fiber-based imaging spectrometer was developed to provide snapshot hyperspectral imaging tuned for biomedical applications. The system is designed for imaging in the visible spectral range from 400 to 700 nm for compatibility with molecular imaging applications as well as satellite and remote sensing. An 81×96  pixel spatial sampling density is achieved by using a custom-made fiber-optic bundle. The design considerations and fabrication aspects of the fiber bundle and imaging spectrometer are described in detail. Through the custom fiber bundle, the image of a scene of interest is collected and divided into discrete spatial groups, with spaces generated in between groups for spectral dispersion. This reorganized image is scaled down by an image taper for compatibility with following optical elements, dispersed by a prism, and is finally acquired by a CCD camera. To obtain an (x,y,λ) datacube from the snapshot measurement, a spectral calibration algorithm is executed for reconstruction of the spatial–spectral signatures of the observed scene. System characterization of throughput, resolution, and crosstalk was performed. Preliminary results illustrating changes in oxygen-saturation in an occluded human finger are presented to demonstrate the system’s capabilities.

Accurate estimation or retrieval of surface emissivity from long-wave infrared or thermal infrared (TIR) hyperspectral imaging data acquired by airborne or spaceborne sensors is necessary for many scientific and defense applications. This process consists of two interwoven steps: atmospheric compensation and temperature–emissivity separation (TES). The most widely used TES algorithms for hyperspectral imaging data assume that the emissivity spectra for solids are smooth compared to the atmospheric transmission function. We develop a model to explain and evaluate the performance of TES algorithms using a smoothing approach. Based on this model, we identify three sources of error: the smoothing error of the emissivity spectrum, the emissivity error from using the incorrect temperature, and the errors caused by sensor noise. For each TES smoothing technique, we analyze the bias and variability of the temperature errors, which translate to emissivity errors. The performance model explains how the errors interact to generate temperature errors. Since we assume exact knowledge of the atmosphere, the presented results provide an upper bound on the performance of TES algorithms based on the smoothness assumption.

The Gough Map, one of the earliest surviving maps of Britain, was created and extensively revised over the 15th century. In 2015, the map was imaged using a hyperspectral imaging system while in the collection at the Bodleian Library, Oxford University. The goal of the collection of the hyperspectral image (HSI) of the Gough Map was to address questions such as enhancement of faded text for reading and analysis of the pigments used during its creation and revision. In particular, pigment analysis of the Gough Map will help historians understand the material diversity of its composition and potentially the timeline of, and methods used in, the creation and revision of the map. Multiple analysis methods are presented to analyze a particular pigment in the Gough Map with an emphasis on understanding the within-material diversity, i.e., the number and spatial layout of distinct red pigments. One approach for understanding the number of distinct materials in a scene (i.e., endmember selection and dimensionality estimation) is the Gram matrix approach. Here, this method is used to study the within-material differences of pigments in the map with common visual color. The application is a pigment analysis tool that extracts visually common pixels (here, the red pigments) from the Gough Map and estimates the material diversity of the pixels. Results show that the Gough Map is composed of at least five kinds of dominant red pigments with a particular spatial pattern. This research provides a useful tool for historical geographers and cartographic historians to analyze the material diversity of HSI of cultural heritage artifacts.

Imaging spectrometers are frequently used in remote sensing for their increased target discrimination capabilities over conventional imaging. Increasing the spectral resolution of these sensors further enables the system’s ability to discriminate certain targets and adds the potential for monitoring narrow-line spectral features. We describe a high spectral resolution (Δλ=1.1  nm full-width at half maximum) snapshot imaging spectrometer capable of distinguishing two narrowly separated bands in the red-visible spectrum. A theoretical model is provided to detail the first polarization grating-based spatial heterodyning of a Savart plate interferometer. Following this discussion, the experimental conditions of the narrow-line imaging spectrometer (NLIS) are provided. Finally, calibration and target identification methods are applied and quantified. Ultimately it is demonstrated that in a full spectral acquisition the NLIS sensor is capable of less than 3.5% error in reconstruction. Additionally, it is demonstrated that neural networks provide greater than 99% reduction in crosstalk when compared to pseudoinversion and expectation maximization in single target identification.

The calculation, interpretation, and implications of radiometric sensitivity metrics for Earth-observing multispectral and hyperspectral imaging sensors are discussed. The most commonly used sensor performance metric is signal-to-noise ratio, from which additional noise equivalent quantities can be computed, including noise equivalent spectral radiance (NESR), noise equivalent delta reflectance (NEΔρ), noise equivalent delta emittance (NEΔϵ), and noise equivalent delta temperature (NEΔT). For hyperspectral sensors, these metrics are typically calculated from an at-aperture radiance (typically generated by MODTRAN) that includes both target radiance and nontarget (atmosphere and background) radiance. Unfortunately, these calculations treat the entire at-aperture radiance as the desired signal, even when the target radiance is only a fraction of the total (such as when sensing through a long or optically dense atmospheric path). To overcome this limitation, an augmented set of metrics based on a contrast signal-to-noise ratio, including their noise equivalent counterparts (CNESR, CNEΔρ, CNEΔϵ, and CNEΔT), is developed. These contrast metrics better quantify sensor performance in an operational environment that includes remote sensing through the atmosphere.

In hyperspectral target detection, one must contend with variability in both target materials and background clutter. While most algorithms focus on the background clutter, there are some materials for which there is substantial variability in the signatures of the target. When multiple signatures can be used to describe a target material, subspace detectors are often the detection algorithm of choice. However, as the number of variable target spectra increases, so does the size of the target subspace spanned by these spectra, which in turn increases the number of false alarms. Here, we propose a modification to this approach, wherein the target subspace is instead a constrained subspace, or a simplex without the sum-to-one constraint. We derive the simplex adaptive matched filter (simplex AMF) and the simplex adaptive cosine estimator (simplex ACE), which are constrained basis adaptations of the traditional subspace AMF and subspace ACE detectors. We present results using simplex AMF and simplex ACE for variable targets, and compare their performances against their subspace counterparts. Our primary interest is in the simplex ACE detector, and as such, the experiments herein seek to evaluate the robustness of simplex ACE, with simplex AMF included for comparison. Results are shown on hyperspectral images using both implanted and ground-truthed targets, and demonstrate the robustness of simplex ACE to target variability.

We study the generalization and scalability behavior of a deep belief network (DBN) applied to a challenging long-wave infrared hyperspectral dataset, consisting of radiance from several manmade and natural materials within a fixed site located 500 m from an observation tower. The collections cover multiple full diurnal cycles and include different atmospheric conditions. Using complementary priors, a DBN uses a greedy algorithm that can learn deep, directed belief networks one layer at a time and has two layers form to provide undirected associative memory. The greedy algorithm initializes a slower learning procedure, which fine-tunes the weights, using a contrastive version of the wake-sleep algorithm. After fine-tuning, a network with three hidden layers forms a very good generative model of the joint distribution of spectral data and their labels, despite significant data variability between and within classes due to environmental and temperature variation occurring within and between full diurnal cycles. We argue, however, that more questions than answers are raised regarding the generalization capacity of these deep nets through experiments aimed at investigating their training and augmented learning behavior.

Drill cores provide direct evidence for geological prospecting, and compiling drill core record information has been a vital tool in geological exploration. We design and fabricate a drill core spectral imaging system that covers a wavelength range of 400 to 2500 nm with a spectral resolution of 3 to 12 nm. A visible and near infrared and a shortwave infrared spectrometer are mounted together to cover the wavelength range. This paper describes the mechanical structure, light path structure, parameters, and software processing chain of the system. With the spectral imaging system and custom-design software, a mineral core sample is analyzed to validate system performance.

This paper discusses computational modeling and experimental demonstration of a Fresnel zone light field spectral imaging system. This type of system couples an axial dispersion binary diffractive optic with a light field detector design providing a snapshot spectral imaging capability. The computational model was validated experimentally and provides excellent predictions of potential system capabilities. Additionally, the experimentally demonstrated prototype was able to digitally refocus monochromatic images by wavelength across greater than a 100 nm bandwidth. Through simulation, the demonstrated system was approximated to have a full range from ∼400 to 800 nm at close to a 15-nm spectral sampling interval. We also demonstrated experimentally the capability of resolving between and processing two different spectral signatures in a single snapshot. The type of system demonstrated here offers substantial new capability as an optically simple, snapshot spectral imager. The experimental proof of concept and computational model set the stage for further development of these types of systems.

Results of a method of estimating index of refraction from passive, polarimetric hyperspectral imaging radiance measurements are presented. As off-nadir viewing hyperspectral imaging platforms gain prominence, estimating index of refraction, which is invariant to viewing angle, may prove advantageous to estimating the emissivity, which is not. Results show that index of refraction can be retrieved to within 8% rms error for fused silica and sapphire glass targets, while simultaneously estimating object temperature. The accuracy and self-consistency of this technique for estimating index of refraction are shown to compare favorably to the maximum smoothness temperature–emissivity separation algorithm. Additionally, the results show that atmospheric downwelling radiance can also be accurately estimated, to within the noise of the instrument, concurrently with index of refraction.

An automatic fishing net detection and recognition method for underwater obstacle avoidance is proposed. In the method, optical gated viewing technology is utilized to obtain high-resolution fishing net images and extend detection distance by suppressing water backscattering and background noise. The fishing net recognition is based on the proposed histograms of slope lines (HSLs) descriptors plus a support vector machine classifier. The extraction of HSL descriptors includes five steps of contrast-limited adaptive histogram equalization, the Gaussian low-pass filtering, the Canny detection, the Hough transform, and weighted vote. In the proof experiments, the detection distance of the fishing net reaches 5.7 attenuation length and the recognition accuracy reaches 93.79%.

A feasible way to improve the manufacturing efficiency of large reaction-bonded silicon carbide optics is to increase the processing accuracy in the ground stage before polishing, which requires high accuracy metrology. A swing arm profilometer (SAP) has been used to measure large optics during the ground stage. A method has been developed for improving the measurement accuracy of SAP using a capacitive probe and implementing calibrations. The experimental result compared with the interferometer test shows the accuracy of 0.068  μm in root-mean-square (RMS) and maps in 37 low-order Zernike terms show accuracy of 0.048  μm RMS, which shows a powerful capability to provide a major input in high-precision grinding.

In the optical stabilization control system (OSCS) control system based on a charge-coupled device (CCD), stabilization performance of the line-of-sight is severely limited by the mechanical resonance and the low sampling rate of the CCD. An approach to improve the stabilization performance of the OSCS control system with load restriction based on three loops, including an acceleration loop, a virtual velocity loop, and a position loop, by using MEMS accelerometers and a CCD is proposed. The velocity signal is obtained by accelerators instead of gyro sensors. Its advantages are low power, low cost, small size, and wide measuring range. A detailed analysis is provided to show how to design the virtual velocity loop and correct virtual velocity loop drift. Experimental results show that the proposed multiloop feedback control method with virtual velocity loop in which the disturbance suppression performance is better than that of the dual loop control with only an acceleration loop and a position loop at low frequency.

We propose an optical thresholder having a flat power transfer function both in low- and high-level outputs. The optical thresholder consists of two cascaded Mach–Zehnder interferometers (MZIs) with a nonlinear microring resonator in each MZI. We theoretically analyze the power transfer function and the response speed of the thresholder made of silicon waveguides. A thresholder having the threshold of 7 mW with a response speed of 13 GHz was obtained with a ring radius of 5  μm. The response speed can be increased to 23.7 GHz by adjusting device parameters with the same ring radius at the cost of increased threshold of 22 mW. Further increased response frequency can be realized by reducing the ring radius. A thresholder with a ring radius of 2  μm shows the response frequency of 40.5 GHz.