Comparing mapping and direct hyperspectral imaging in stand‐off Raman spectroscopy for remote material identification

• Published in: Journal of Raman spectroscopy Vol. 50; no. 7; pp. 1034 - 1043
• Main Authors: Gasser, Christoph, González‐Cabrera, María, Ayora‐Cañada, María José, Domínguez‐Vidal, Ana, Lendl, Bernhard
• Format: Journal Article
• Language: English
• Published: Bognor Regis Wiley Subscription Services, Inc 01.07.2019; John Wiley and Sons Inc
• Subjects: Algorithms; Chemical composition; Color imagery; Field of view; Hyperspectral imaging; Image processing; Laser beams; Light; Liquid crystals; Mapping; Noise; Organic chemistry; Polyimide resins; Raman spectroscopy; remote detection; Spectra; Spectral resolution; Spectroscopy; Spectrum analysis; stand‐off; Tunable filters
• ISSN: 0377-0486; 1097-4555
• DOI: 10.1002/jrs.5607

Stand‐off Raman spectroscopy offers a highly selective technique to probe unknown substances from a safe distance. Often, it is necessary to scan large areas of interest. This can be done by pointwise imaging (PI), that is, spectra are sequentially acquired from an array of points over the region of interest (point‐by‐point mapping). Alternatively, in this paper a direct hyperspectral Raman imager is presented, where a defocused laser beam illuminates a wide area of the sample and the Raman scattered light is collected from the whole field of view (FOV) at once as a spectral snapshot filtered by a liquid crystal tunable filter to select a specific Raman shift. Both techniques are compared in terms of achievable FOV, spectral resolution, signal‐to‐noise performance, and time consumption during a measurement at stand‐off distance of 15 m. The HSRI showed superior spectral resolution and signal‐to‐noise ratio, while more than doubling the FOV of the PI at laser power densities reduced by a factor of 277 at the target. Further, the output hyperspectral image data cube can be processed with state of the art chemometric algorithms like vertex component analysis in order to get a simple deterministic false color image showing the chemical composition of the target. This is shown for an artificial polymer sample, measured at a distance of 15 m. We present comparison between a mapping stand‐off Raman imager and a direct hyperspectral Raman imager (HSRI) employing a tunable filter. The performance of both imagers is contrasted regarding spectral and spatial resolution, signal to noise, and measurement time. The importance of having high spatial resolution for large area scanning of trace amounts is highlighted. Image processing using chemometrics with the acquired HSRI images is shown.