A low-cost hyperspectral scanner for natural imaging and the study of animal colour vision above and under water

Abstract

Hyperspectral imaging is a widely used technology for industrial and scientific purposes, but the high cost and large size of commercial setups have made them impractical for most basic research. Here, we designed and implemented a fully open source and low-cost hyperspectral scanner based on a commercial spectrometer coupled to custom optical, mechanical and electronic components. We demonstrate our scanner’s utility for natural imaging in both terrestrial and underwater environments. Our design provides sub-nm spectral resolution between 350–950 nm, including the UV part of the light spectrum which has been mostly absent from commercial solutions and previous natural imaging studies. By comparing the full light spectra from natural scenes to the spectral sensitivity of animals, we show how our system can be used to identify subtle variations in chromatic details detectable by different species. In addition, we have created an open access database for hyperspectral datasets collected from natural scenes in the UK and India. Together with comprehensive online build- and use-instructions, our setup provides an inexpensive and customisable solution to gather and share hyperspectral imaging data.

What the Zebrafish’s Eye Tells the Zebrafish’s Brain: Retinal Ganglion Cells for Prey Capture and Colour Vision

“…A segregation of achromatic mid/long-wavelength vision and circuits for colour vision is arguably taken to the extreme in the eyes of many arthropods including fruit flies: Six of each ommatidium's eight photoreceptors (R1-6) express the same mid-wavelength opsin to support achromatic image forming vision, while the remaining two photoreceptors (R7,8) in parallel provide information about contrasts in wavelength (Heath et al, 2020;Schnaitmann et al, 2018). Similarly, like many visual neurons in insects (Chen et al, 2019;Heath et al, 2020;Schnaitmann et al, 2018;Yang et al, 2004), the finding that also in zebrafish most opponent RGCs encode simple rather than complex opponencies is in line with previous work (Baden and Osorio, 2019;Kamermans et al, 1991Kamermans et al, , 1998Zimmermann et al, 2018) and links to the predominance of simple-over complex spectral contrasts in natural scenes (Buchsbaum and Gottschalk, 1983;Lewis and Zhaoping, 2006;Maloney, 1986;Nevala and Baden, 2019;Ruderman et al, 1998;Zimmermann et al, 2018).…”

“…Similarly, the dominance of long-over short-wavelength responses in the lower visual field (Fig. 3D) is likely related to the predominance of long-wavelength light in the lower water column (Muaddi and Jamal, 1991;Nevala and Baden, 2019), as well as the zebrafish's behavioural need to monitor the ground beneath them for systematic image shifts that drive a long-wavelength biased optomotor response (Orger and Baier, 2005;Wang et al, 2020).…”

“…For stimulation, we presented full-field light modulated in time and wavelength based on four LEDs that were spectrally aligned with the sensitivity peaks of the zebrafish's four cone opsins (R, G, B, UV) (Zimmermann et al, 2018). The power of each LED was adjusted to follow the relative power distribution across wavelength of daytime light in the zebrafish natural habitat (Nevala and Baden, 2019;Zimmermann et al, 2018) to yield a "natural white": red (100%), green (50%), blue (13%) and UV (6%) (Fig. 2B).…”

“…1. It does not explain why nearly half of all output channels are colour opponent it should be substantially fewer (Lewis and Zhaoping, 2006;Nevala and Baden, 2019 One explanation for this series of apparent mismatches with dominant theories in colour vision might relate to an implicit assumption that most if not all spectral processing and opponency should in some way link to image forming colour vision (Baden and Osorio, 2019). However, spectral information can be useful in additional ways.…”

“…First, zebrafish have a large field of view that lets them simultaneously survey the overhead sky and the riverbed beneath them (Bianco et al, 2011;Patterson et al, 2013;Zimmermann et al, 2018). These parts of visual space have vastly different behavioural relevance, as well as distinct spatial, temporal and spectral statistics (Baden et al, 2020;Engeszer et al, 2007;Nevala and Baden, 2019;Parichy, 2015;Zimmermann et al, 2018). For efficient coding (Attneave, 1954;Barlow, 1961;Simoncelli and Olshausen, 2001) zebrafish should therefore invest in different sets of functional RGC types to support different aspects of vision across their retinal surface.…”