Hyperspectral imaging

Reflectance imaging spectroscopy (IS) records the radiation reflected from the surface of an object and captures the reflectance spectrum for each pixel in the image. It is an imaging technique that sequentially captures hundreds of reflectance frames over narrow, nearly contiguous spectral bands. When the spectral resolution is only a few nm, IS is referred to as hyperspectral imaging (HSI).

HSI represents a very powerful spectroscopic technique in which images are taken over a wide wavelength range, and can be used for high-quality documentation of the work. A hyperspectral camera can simultaneously acquire an image and, for each pixel, the corresponding spectrum. By combining imaging and spectroscopy, this generates what is commonly called a 'data cube'.

The latest frontier of hyperspectral devices lies in the availability of a wide range of compact, portable and user-friendly devices. A Specim IQ hyperspectral camera (Specim, Spectral Imaging Ltd., Oulu, Finland) was used in the present study. The camera operates in the 400–1000 nm spectral range (using a silicon CMOS sensor) with a spectral resolution of about 7 nm, and records 204 spectral bands on 512×512 pixels. The novelty of the device lies in its reduced dimensions (207×91×74 mm) and weight (1.3 kg), and in its user-friendly interface, which make this camera much more portable, manageable and affordable than conventional hyperspectral instrumentation. The optical module of the camera makes it possible to image areas of different size, by varying the working distance from short distances (tens of centimeters) to long distances (tens of meters) with different spatial resolutions.

The recommended sources for the Specim IQ are halogen sources, which cover the entire sensitivity range of the camera (www.specim.fi). During the present study, two 12 V, 50 W halogen lamps from Osram were used. The lamps were mounted symmetrically at 45° to the normal of the surface to be acquired, in order to illuminate it uniformly. These low-intensity sources were selected to ensure minimal impact on a highly photosensitive material such as parchment, and the amount of radiation sent onto the codex during acquisition was measured using a data logger. The values were approximately 2000 lux (lm/m²) in the visible and 45 mW/m² in the UV-A, with a UV-A/visible ratio of 22.5 μW/lm, the latter being below the maximum acceptable UV-A radiation dose.

The device is based on the use of a next-generation imaging spectrograph with a diffractive transmissive component. Acquisition over the investigated area is performed in push-broom mode, i.e. by line scanning with an integrated scanner. The camera is also equipped with a user interface system to allow real-time control over data acquisition, processing and visualization.

White target calibration was performed at the beginning and end of the measurement session, using the certified Spectralon® 99% diffuse reflectance white reference supplied with the camera. The white target is usually framed within the scene being investigated. However, the system also supports an operating mode to acquire the white reference separately.

The data were processed using the Specim IQ Studio PC software, and a mapping of the different materials was performed using the Spectral Angle Mapper (SAM) algorithm. This method evaluates the similarity between a reference spectrum and the spectra associated with the pixels of the acquired data. All the pixels with the same spectral behavior can then be visualized in false-color images; this made it possible to map the inhomogeneities of the parchment, as well as the distribution of the pigments and inks used.

Hyperspectral imaging setup at the Biblioteca Nazionale Vittorio Emanuele III of Naples.

References

1. Cucci, C., Casini, A., Stefani, L., Picollo, M. and Jussila, J., 2017. Bridging research with innovative products: a compact hyperspectral camera for investigating artworks: a feasibility study. In: Optics for Arts, Architecture, and Archaeology VI (Vol. 10331, pp. 17-29). SPIE.
2. Cucci, C. and Casini, A., 2019. Hyperspectral imaging for artworks investigation. In: Data Handling in Science and Technology (Vol. 32, pp. 583-604). Elsevier.
3. Picollo, M., Bacci, M., Casini, A., Lotti, F., Poggesi, M. and Stefani, L., 2007. Hyperspectral image spectroscopy: a 2D approach to the investigation of polychrome surfaces. Proceedings of Conservation Science, pp.10-11.
4. Picollo, M., Cucci, C., Casini, A. and Stefani, L., 2020. Hyper-spectral imaging technique in the cultural heritage field: New possible scenarios. Sensors, 20(10), p.2843.
5. Picollo, M., Casini, A., Cucci, C., Stefani, L., Jiménez-Garnica, R. and Fuster-López, L., 2020. Documentation and analysis of some Picasso's paintings by using hyperspectral imaging technique to support their conservation and stylistic matters. In: IOP Conference Series: Materials Science and Engineering (Vol. 949, No. 1, p. 012023). IOP Publishing.

Keywords: Hyperspectral, Imaging