Instrumental systematic errors in a chromotomographic hyperspectral imaging system

Abstract

Chromotomographic imaging (CTI) is a method of convolving spatial and spectral information which can be reconstructed into a 3-D spatial/spectral image cube using the same transforms employed in medical tomography. The Air Force Institute of Technology (AFIT) has built a chromotomographic imaging system which features a rotating prism to accomplish this convolution. This paper will specifically address effects of systematic instrumental error in collected projection data and on the reconstructed hyperspectral image cube in terms of spatial location and resolution, spectral line shift and increased width of spectral peaks. Systematic error considered are those associated with prism alignment, detector array position, and prism rotation angle. The methods used to reconstruct hyperspectral data cubes require an approximation of the instrumentation which generates the forward transform. Any variation between the estimate of these properties and their true performance results in a residual term in the reconstruction of a spectral bin, which is characterized as an error kernel in the transform equation. A computer model will be used to explore the sensitivity of the kernel to various systematic errors. Data is then collected with the instrument to assess the validity of the model under some representative conditions. It is found that the shorter wavelength (λ≪500nm) 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≫600nm) suffer mostly from a shift of the spectral peaks. The model as well as collected data indicates that 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 from 50%–100% in the blue spectral region. For larger errors than this, spectral peaks begin to split into bimodal distributions, and spatial point response functions (PRFs) are reconstructed in rings with radius proportional to wavelength and thickness of the PRF. 1,2