Hyperspectral Camera and Its Application for Plant Phenotyping and Disease Detection

Use of Specim IQ Camera System as a Tool for Understanding Physiological Response of Arabidopsis thaliana Mutants Adapted in Stressed and Non-Stressed Condition

Vegetation indices (VIs) are used by plant scientists to evaluate structural and functional vegetation traits quantitatively and qualitatively across different scales [24,25,51,52]. They are used for example to assess the green biomass [53], canopy structure [54], leaf area index [55], chlorophyll content [56,57,58], fraction of absorbed photosynthetically active radiation (fAPAR) [55,59], and light-use efficiency [29] of plants and canopies [60]. With the development of portable hyperspectral imaging sensors, reflectance from visible (400–700 nm) and near infrared (700–1000 nm) spectrum can be easily derived to study and screen mutant plants to identify genes and understand its physiological function in a high-throughput way in phenotyping environments.
In this case study, two mutant lines of Arabidopsis thaliana acclimated in non-stressful (NSA) and stressful (SA) condition were used as the main subject. The selected mutants are deficient in either PsbS protein (npq4) [61] or violaxanthin de-epoxidase (npq1) [62], both resulting in a limited ability to thermally dissipate excessive light energy via a process called non-photochemical quenching (NPQ). In addition, the lack of violaxanthin de-epoxidase in npq1 mutant inhibits light-dependent enzymatic conversion of carotenoids. This conversion, which is a part of NPQ regulation, causes a very subtle spectral change in leaf reflectance that is invisible to human eyes. Both mutations are not fatal and the plants develop normally under greenhouse conditions. Under high light conditions, however, the two mutants are unable to adjust light energy harvesting by NPQ to protect their photosystems against photo-damage. Along with these mutants, the “normal” plants without mutations (Col-0) were used as a control group. Thus, by using these mutants we tested whether the Specim IQ can detect subtle differences in leaf pigments and physiological traits.
Plants were sown in 7 × 7 cm pots (one plant per pot) filled with soil. Three plants from each genotype were grown and acclimated to non-stressful conditions in the greenhouse while another three plants were acclimated for at least two days to natural light and temperature conditions in the field that are more stressful for plants. All plants were randomly distributed within the imaging frame.
The Specim IQ camera system was used to take reflectance images of Arabidopsis plants inside the greenhouse with white panel (90% reflectance) as a reference target. As the Specim IQ software does not support the calculation of vegetation indices, the hyperspectral imagery data was imported into ENVI Classic 5.3 (Harris Geospatial Solutions, Broomfield, USA) resolving the regions of interest (ROI) using Normalized Difference Vegetation Index (NDVI [63], Equation (1)) and manual tracing of individual plants. For the case study, two VIs were calculated that are correlated to leaf chlorophyll content, namely the NDVI and the Red Edge Inflection Point (REIP [64,65], Equation (2) by Matlab 2013a and Signal Processing Toolbox 6.19 (The MathWorks, Inc., Natick, USA)). To derive REIP, plant spectra was smoothed using Savitzky–Golay filtering [49] before calculating the first derivative. REIP was identified as the maximum value of the first derivative between 690 and 720 nm after spline interpolation with 0.1 nm resolution. As a third VI, the Photochemical Reflectance Index (PRI [66], Equation (3)) was calculated, which is described to be sensitive to the activation of the NPQ pathway and carotenoid conversion (for more details on these indices and their functional meaning see [25]. Using the RStudio software V1.0.143 (RStudio, Inc., Boston, USA), analysis of variance (ANOVA) was carried out using agricolae package for all the calculated average VIs in a single plant. Likewise, pairwise mean comparison using least square difference (LSD, α=0.05) was performed as a post hoc analyses.

Laisser un commentaire