Paper Title
HYPERSPECTRAL IMAGING RETRIEVAL IN THE VIS-ESWIR BANDS USING A NON-PARAMETRIC APPROACH FOR TURBULENT FLAMES

Abstract
Hyperspectral imaging (HSI) has become a promising technique due to its potential to capture spatial-spectral-temporal information that feeds classification and/or estimation algorithms. However, current optical configurations make systems more complex and expensive, making them impractical for real-time applications where fast sensor responses and high spatial-spectral resolutions are needed. In this paper, we propose a novel HSI based on a high-spatial-resolution VIS-eSWIR (from visible to extended short-wave-infrared bands) monochromatic camera, a set of only six selected broadband optical filters, and a non-iterative-hyperspectral-retrieval-technique to estimate spatial-spectral-temporal data from turbulent flames. The proposed method is wholly based on apriori-measured information from the emitted flame spectra. To acquire spectra, two previously calibrated spectrophotometers (one for the VIS band and the other for the SWIR band) were used, thus providing a learning spectral matrix of 180 samples of 562 channels. Thus, the hyperspectral estimated data involves the range 400nm - 2100nm and a spectral resolution of about 3nm. The effectiveness of the spectral retrieval procedure was compared with the learning spectra using the GFC (goodness-of-fit coefficient) and the RMSE (root-means square error) metrics, providing a mean GFC of 0.9999 and a mean RMSE of 0.0042. With such results, we demonstrate the effectiveness of the implemented HSI system, thus providing a novel, fast, and less complex spectral tool for process monitoring. Keywords - Hyperspectral retrieval, Imaging, Flames