The spatial and spectral resolutions achievable by a prototype rotating prism chromotomosynthetic imaging (CTI) system operating in the visible spectrum are described. The instrument creates hyperspectral imagery by collecting a set of 2D images with each spectrally projected at a different rotation angle of the prism. Mathematical reconstruction techniques that have been well tested in the field of medical physics are used to reconstruct the data to produce the 3D hyperspectral image. The instrument operates with a 100 mm focusing lens in the spectral range of 400-900 nm with a field of view of 71.6 mrad and angular resolution of 0.8-1.6 μrad. The spectral resolution is 0.6 nm at the shortest wavelengths, degrading to over 10 nm at the longest wavelengths. Measurements using a point-like target show that performance is limited by chromatic aberration. The system model is slightly inaccurate due to poor estimation of detector spatial resolution, this is corrected based on results improving model performance. As with traditional dispersion technology, calibration of the transformed wavelength axis is required, though with this technology calibration improves both spectral and spatial resolution. While this prototype does not operate at high speeds, components exist which will allow for CTI systems to generate hyperspectral video imagery at rates greater than 100 Hz.

Chromotomosynthetic imaging (CTI) is a method of convolving spatial and spectral information that can be reconstructed into a hyperspectral image cube using the same transforms employed in medical tomosynthesis. A direct vision prism instrument operating in the visible (400-725 nm) with 0.6 mrad instantaneous field of view (IFOV) and 0.6-10 nm spectral resolution has been constructed and characterized. Reconstruction of hyperspectral data cubes requires an estimation of the instrument component properties that define the forward transform. We analyze the systematic instrumental error in collected projection data resulting from prism spectral dispersion, prism alignment, detector array position, and prism rotation angle. The shifting and broadening of both the spectral lineshape function and the spatial point spread function in the reconstructed hyperspectral imagery is compared with experimental results for monochromatic point sources. The shorter wavelength (λ<500 nm) region where the prism has the highest spectral dispersion suffers mostly from degradation of spectral resolution in the presence of systematic error, while longer wavelengths (λ>600 nm) suffer mostly from a shift of the spectral peaks. The quality of the reconstructed hyperspectral imagery is most sensitive to the misalignment of the prism rotation mount. With less than 1° total angular error in the two axes of freedom, spectral resolution was degraded by as much as a factor of 2 in the blue spectral region. For larger errors than this, spectral peaks begin to split into bimodal distributions, and spatial point response functions are reconstructed in rings with radii proportional to wavelength and spatial resolution.

Numerous hyperspectral algorithms have been developed to detect both full and sub-pixel solid target materials. Target signatures are obtained from spectral libraries that contain both target and non-target materials. When the library is large and contains many potential targets, it is inefficient to run an individual detector for each material of interest. Additionally, such an approach produces numerous false alarms (i.e., multiple detections per pixel) due to spectral similarity among targets. In this paper, we present an efficient approach for detecting multiple targets within large spectral libraries while mitigating false alarms. We first group spectrally similar materials within the library into a hierarchy of clusters. From each cluster containing a target material, a single detector is obtained. Each detector represents multiple library spectra, so an identification step is needed for detected pixels. Detected pixels are modeled as a mixture between their local in-scene background and candidate library spectra. Candidates are chosen from adjacent library clusters. The candidate model providing the best fit is chosen to report. Use of local background spectra provides a physically meaningful mixing model that adapts to detected pixels. Clustering the library reduces the computational complexity of modeling detected pixels. We demonstrate detection and false alarm mitigation performance of our proposed algorithm with a real hyperspectral dataset.