Frost is estimated to cost Australian grain growers $\$$ 360 million in direct and indirect losses every year. Assessing frost damage manually in barley is labor intensive and involves destructive sampling. To mitigate against significant economic loss, it is crucial that assessment decisions on whether to cut for hay or continue to harvest are made soon after frost damage has occurred. In this paper, we propose a non-destructive technique by using raster-scan terahertz imaging. Terahertz waves can penetrate the spike to determine differences between frosted and unfrosted grains. With terahertz raster-scan imaging, conducted in both transmission and reflection at 275 GHz, frosted and unfrosted barley spikes show significant differences. In addition, terahertz imaging allows to determine individual grain positions. The emergence of compact terahertz sources and cameras would enable field deployment of terahertz non-destructive inspection for early frost damage.
© 2020 Optical Society of America under the terms of the OSA Open Access Publishing Agreement
Wheat and barley are two of the most important temperate cereals cultivated around the world [1,2], in part due to their broad ecological adaptation to many varying environments. Wheat is used mainly to meet human food demands, while barley—based on the volume of production—is mainly used for animal feed and secondly for malting .
Grain yield is largely determined by the number of grains per square meter in bread wheat [4–10], durum wheat [11–13], barley  and other grain crops [15,16]. The conceptual model generally used to understand the physiology of grain number identifies critical periods within the crop cycle when the amount of available assimilates limits growth and yield . For grain number in wheat, this period most frequently starts with the emergence of the penultimate leaf at 20–30 days prior to anthesis and ends at around 7-10 days after flowering (anthesis, DC6.5 [17,18]). In two-row barley, this critical period is reported between 40 and 10 days before the heading stage .
In Mediterranean regions such as southern Australia, wheat and barley crops are frequently exposed to abiotic stress conditions such as water deficiency or high temperature (with hot northerly winds) that increase in intensity during the grain filling phase of the growing season, leading to reduced yield and grain quality . To avoid this, grain growers may sow their crops earlier in the season. However, in doing so this shifts flowering time back and exposes the plants to low temperature and frost damage during late winter and early spring. Frost damage is a result of the reproductive organs being exposed to a wide range of frost severity (i.e., intensity and duration of each frost event), and their occurrence during the growing season with air temperature values that are recorded at canopy level below 0°C .
Therefore, depending on the genotype selection (G), management practices (M), environmental conditions (E) and their interactions (G×M×E), there is a wide range of susceptibility to frost damage in cereals that increases sharply after booting [22–24]. One-degree difference in temperature could result in frost damage escalating from 10% to 90% in wheat , resulting in economic losses in Australia around $\$$ 360 million every year [26,27].
The current National Frost Initiative led by The University of Adelaide has been benchmarking wheat and barley varieties across multiple years, sites and times of sowing (ToS) to rank radiant frost susceptibility in excess of one thousand spikes per year based on empirical manual determinations. However, analyzing the fate of floret primordia that are becoming grains is far more difficult and time-consuming than determining tiller numbers at different times during the season . In barley, floret primordia or grains within the spikelet are protected by two bracts, one external—the lemma, and one internal—the palea, making it difficult to quantify frost damage without removing them. Ideally, plant breeders are constantly seeking new methods and/or cutting edge tools/platforms that will enable them to phenotype large numbers of genotypes quickly and accurately that have been exposed to different abiotic stress conditions. Ultimately, access to such developments increases breeding efficiency .
Most of the recent studies of reproductive frost damage in wheat include: (i) the utility of hyperspectral reflectance and active light fluorescence , (ii) spectral mixture analysis under field conditions and semi-controlled environments [31,32], and (iii) X-ray tomography  across different environmental conditions from field to controlled environments, respectively, Moreover, (iv) thermal imaging with remote sensing or field-based high-throughput phenotyping platforms has also been investigated and employed in several studies [29,32]. In addition, a recent study in drought and heat stress tolerance in wheat using X-ray computed tomography could be an alternative approach to evaluate frost damage in cereals, albeit emitting harmful radiation to users of the technology . To date, there are many challenges associated with non-destructive methodology for evaluating frost damage feasibly and accurately. In addition, the practicalities for grain growers to quickly determine which crop should be cut for hay or left to grow for harvest in the field is important.
As an alternative non-destructive approach, terahertz imaging is non-ionizing and can penetrate most dry and non-metallic materials. These benefits are promising for applications in agriculture . Previous studies using terahertz technology in agriculture include seed classification [36,37], rice seed discrimination , identification of soybean varieties , evaluating plumpness of sunflower seeds , quality of seeds  and identification of seed infestation in sheep . Particularly, terahertz spectroscopy has been used to determine grain quality in wheat by combining information regarding the absorption and refractive index of wheat in various conditions [43,44]. Apart from that, terahertz spectroscopy has been applied to crushed wheat to determine moisture content . Such results have shown that terahertz spectroscopy, combined with mathematical models is applicable for determining wheat quality with good accuracy. Currently, there are several existing methods for terahertz imaging techniques including full-field imaging , confocal imaging  and single-pixel imaging  which can provide rapid image acquisition in place of terahertz spectroscopy. In this study, we provide proof-of-concept that illustrates how terahertz imaging can be used to assess frost damage in barley. For reflection imaging, time-domain background subtraction was utilized to remove cluttering reflections. This technique provided additional information in the images of the spikes with respect to grain water content. Additionally, we corroborate empirical measurements of frost damage using terahertz imaging to make economic decisions with regards to the plant.
2.1 Experimental field conditions
One experiment was carried out under frost-prone field conditions in Loxton (34°28'10.03"S, 140°34'3"E, 37 m.a.s.l.), South Australia in 2019 as part of The National Frost Initiative Nursery led by the University of Adelaide. Targeted plant density was 180 plants per square meter. Seeds were sown using an experimental planter where each plot size was 6 rows (3-m length) with 26 cm between rows. Weeds, diseases, and insects were prevented by spraying recommended herbicides, fungicides and insecticides at the doses suggested by their manufacturers. This trial included six times of sowing (ToS): 18th April, 29th April, 09th May,' 20th May, 30th May and 13th June 2019. The initial soil nitrogen (0–100 cm depth) was 26 kg N ha-1. At sowing (DC0.0; ) 50 kg ha-1 of Mono-Ammonium Phosphate (MAP; 12% N) was added, while at the beginning of jointing (DC3.1) 60 kg N ha-1 as Urea Ammonium Nitrate (UAN) was applied as a spray formula.
Air minimum and maximum temperature values were recorded from the closest weather station to the experiment (ID:24024, Loxton Research Station, BOM; ). In addition, another weather station was located in the paddock . A wide range of sowing times enabled a wide range of flowering times to be observed, while coinciding with different reproductive frost events from August to September. The lowest recorded temperature value was -5.5°C, (Fig. 1).
2.2 Treatment, experimental design and measurements
Treatments consisted of the combinations of 36 two-row barley varieties sown across six different ToS. Each ToS was assigned to a whole block and each variety was randomized as field plots within sub-blocks, and with two replicates per entry (see details of the experimental design in ). For proof-of-concept, we focused on the comparison of two contrasting ToS: an early sowing date (ES: 18th April) and late sowing (LS: 30th May). This in turn enabled us to evaluate differences in severity and occurrence of the total number of frost events during the critical period for fertile floret and grain number determination around flowering in barley (DC4.9-DC5.2, see Fig. 1, upper panel). In addition to selecting two of the ToS for evaluation, we chose to compare two (Keel and La Trobe) of the 36 two-row barley varieties. These were selected based on (i) their maturity type being similar (thus avoiding any potential differences that could have been caused by phenology), and (ii) their contrasting susceptibility to frost damage (low and high, respectively) recorded during the last seven years at Loxton (South Australia) from the decision making platform: FV-Plus Frost Rankings (Fig. 2, ), generated in collaboration with the University of Wollongong (Centre for Bioinformatics and Biometrics – CBB).
From sowing to physiological maturity, phenology was recorded weekly using the decimal code (DC) from . Anthesis was recorded when 50% of plants from the inner part of each plot (avoiding border effects) reached the end of the booting stage (DC4.9) or just immediately after the beginning of heading (DC5.2). In parallel, manual dissection of the central spikelets of those plants in which the anthers of floret reached pale green to bright yellow in color (fertile florets) following the scale proposed by  was also conducted, as exemplified in [54,55]: see Figs. 1 and 2 therein, respectively.
Within 48 hours following each frost event (accumulated air temperature below 0°C at canopy height, registered every 30 minutes), 30-40 main-shoot stems per plots at the anthesis stage were tagged with a colored ribbon per frost event using a hand operated tagging gun. The tagged spikes were left to develop until the soft dough stage (DC8.5), after which they were individually collected, labeled and bagged. In the laboratory, a random sub-sample of five spikes from each collection were oven-dried (labeled ‘O’) at 40°C for 48h and another five spikes were frozen (labeled ‘F’) at -20°C (conventional freezer) in order to determine whether storage made a difference to the results obtained.
Empirical measurements of frost damage were calculated as the ratio of frosted grains to the total number of grains expressed as a percentage along the spike.
2.3 Terahertz measurements and analysis
To assess frost damage non-destructively, a terahertz electronic system is used to collect complex transmission and reflection data, namely amplitude and phase responses of the barley spikes. The vector network analyzer (VNA) sends a microwave signal through the VDI extension module (VNAX WR-3.4) that emits a terahertz signal in the frequency range of 220 to 330 GHz. This frequency range is selected due to the low absorption from atmospheric water vapor  and the grains themselves. Two horn antennas make up the transmitter and receiver end of the terahertz extension modules and are configured to be vertically polarized. The parameters of the lenses are chosen to complement the beam diameter from the transmitter to generate a small focused spot. As a larger effective aperture enabled focusing of the beam down to a small spot, lenses with a large focal length are selected as the collimating lenses. Subsequently, lenses with a shorter focal length will focus the beam down to a small focused spot. As shown in Fig. 3, the terahertz wave is collimated through a lens with focal length of 200 mm (Lens 1), which is then focused down to a spot size of ∼1.9 mm by a lens with a focal length of 75 mm (Lens 2). A knife-edge test is conducted to determine the actual spot size, which is the practical resolution of the system. The small focused spot is essential for enhanced image resolution of the barley spikes. The data is acquired from the VNA in transmission and reflection simultaneously, as the terahertz extension module is a transceiver. Four barley spikes with different ToS and storage methods are placed horizontally in an acrylic-frame sample holder. It is noteworthy that horizontal polarization does not alter the results obtained. In order to undertake a raster scan, the sample was displaced by two linear translation stages with stepper motors that travel along a 5 cm distance in the x- and y-axes. A raster scan is then performed with 100 steps in the x- and y-direction with a step size of 0.5 mm which takes approximately 26 hours. In the reflection mode, data post-processing is required to remove spurious reflections from the lenses and horn antennas. To do so, inverse Fourier transform was applied to yield the time-domain signals. Time-domain data background subtraction and time-gating was then applied to remove spurious reflections from optical components. For time-domain background subtraction, a time-domain pulse that originated from a known background pixel is subtracted from the rest of the pulses in the terahertz image. With this subtraction, the pulse that remained is primarily due to the barley spikes themselves. Fourier transform is then applied to the time-gated data to recover the frequency response of the sample. All the datasets are consistently processed, and all results shown are taken at the central frequency of the system which is 275 GHz.
3. Results and discussion
Averaging across varieties, empirical measurements of the early sowing time (in the 2019 growing season at Loxton) showed a significant difference in the percentage of frost sterility (92.5 ± 9.6%) compared to the late sowing (7.3 ± 5.1%), largely due to anthesis coinciding with the highest severity and occurrence of frost events resulting in severe damage of fertile florets and grain development compared to delayed flowering by LS (Fig. 1, upper panel). In addition, a wide range of susceptibility to frost damage in spikes was observed through measurements using the terahertz raster scan. Figure 4 shows the transmission and reflection images collected from the La Trobe barley variety. In the first column of Fig. 4, the LS spikes have a lower transmission amplitude when compared to the ES spikes. Correspondingly, from the second column in Fig. 4, the reflection amplitude shows that the grains with the highest reflection amplitude (in red) are more profound in the LS spikes. This high reflection and corresponding low transmission indicate that the LS spikes are unfrosted. Despite that, in the ES samples, terahertz imaging indicate that some grains survived frost damage. It is worth mentioning that even though the plant experiences a frost event and may appear to be frosted, the grains within the spikelet may still be viable. While grain growers are looking for visual symptoms of frost damage during grain filling (milk stage or later), terahertz images can penetrate spikes to accurately detect much earlier signs of floret primordia abortion before anthesis. Regarding the storage, from the first column, the transmission amplitude for the oven-dried LS spikes indicates a higher transmission amplitude when compared to their frozen counterparts.
Figures 5(a)–5(c) depicts the transmission, reflection and optical RGB images of the barley variety Keel across three biological replicates. In the first column, the transmission amplitude for the LS spikes are significantly lower than the ES spikes. This indicates that the LS spikes are unfrosted while the ES spikes are likely to be frosted. It is also noteworthy that the Keel cultivar transmission amplitude is lower than that obtained for the La Trobe cultivar. This is consistent with the fact that the Keel cultivar has a higher tolerance to frost damage when compared to La Trobe. In the second column of Fig. 5, we observe the individual grain positions within the spikelet in the reflection amplitude across the three biological replicates. As these grains are potentially viable for continual growth, they contain a high moisture content, thus having a high reflection amplitude. Figures 5(a)–5(b) shows a healthy presence of grain in both the LS and ES sets while in the second column of Fig. 5(c), the lack of grain in the ES spikes are more profound. However, it is likely that the ES spikes in Fig. 5(c), both O and F are badly frosted.
In the late sowing, where most of the grains were unfrosted (i.e., plants reached anthesis at the end of September avoiding most of the frost events), we compared the magnitude of frost damage between La Trobe and Keel varieties under less stressed conditions. As shown in Fig. 6, the amplitude of the Keel variety was lower in transmission and higher in reflection as compared to the La Trobe variety across all biological replicates. This confirms a lower susceptibility to frost damage in Keel, being in line with the F-Plus frost ranking results from 2012 to 2019 for these varieties in Southern Australia (Fig. 2, ). The terahertz scans were also able to identify the missing grains from the transmission and reflection images as shown by the red circles in Figs. 6(a)–6(b), which are not as clear from the optical images. Thereafter, a germination test was required to determine the viability of those potential seeds. The general trend between ToS was consistent with the predicted values of the terahertz raster scan, where plants sown late (LS) achieved anthesis (FL) in the middle of September and thus avoided most of the frost events, compared to plants sown early (ES) which were exposed to multiple frost events (Fig. 1).
To validate differences between empirical and predicted values, a germination test was conducted with the spike being separated (Rep 3) or unaltered (Rep 2) as shown in Fig. 7. Then, the viability of grains at each spikelet positions along the spike (from the bottom to the tip of the spike) were quantified. The viability of those grains was in line with the higher reflection amplitude as shown in Fig. 7(a), and the empirical values across replicates showed that the probability (expressed as percentage) of having an unfrosted grain within each spikelet position was equal to 100%. Figure 7(c) shows the successful germination test where both replicates 2 and 3 contained grains that were viable throughout the entire spike. To summarize, transmission imaging allowed us to differentiate between frosted and unfrosted grains (Fig. 4 and Fig. 5). In addition, we quantified through a second germination test whether some grains under early-sown conditions survived frost damage which was apparent in the terahertz images (Fig. 8). Only three grains in the basal positions of the spike germinated as shown in Fig. 8(c), which is consistent to the severe early frost events (Fig. 1). Reflection imaging showed the individual grain positions that allowed quantification of viable grain. Lastly, the germination test in accordance to the reflection images confirmed the viability test.
The laboratory set-up employed raster-scanning that concentrates the source power onto each individual pixel. If a camera is used for full-field imaging, the power from the source must be spread across the entire field of view. Typically, the actual power spread follows a Gaussian distribution. Here, a uniform power spread is used to approximate the power required from the source. For practical field applications, reflection imaging is ideal to mimic the field test. The power required by the source can then be determined from the maximum attenuation of the sample, the number of pixels, and the camera sensitivity. The measured maximum attenuation detected from the barley samples is around -42 dB. Assuming a terahertz camera has a sensitivity of -70 dBm and has 10,000 pixels in total, it is estimated that the required power level from the terahertz source is 12 dBm or 15.85 mW at 275 GHz.
In this paper we have explored how terahertz imaging in both transmission and reflection was able to discriminate between frosted and unfrosted barley spikes, beyond frost severity. A raster-scan using terahertz detectors at a center frequency of 275 GHz was shown to be a viable approach. Additionally, reflection mode terahertz imaging enabled quantification of the grain within a spike. A proceeding germination test determining grain viability within the spikelet was consistent with the results obtained from terahertz imaging. The ability to distinguish between frosted and unfrosted spikes is a key point to ensure that the most appropriate decision is being made following frost events. Future research to discriminate quantitatively the thresholds of frost damage in barley and wheat needs to be undertaken. This work can be extended to wheat. However, terahertz imaging of wheat requires a set-up with suitable optics to account for its non-planar spikes. Furthermore, identifying an appropriate terahertz source and camera to create a prototype for practical stand-off imaging applications in the field would significantly improve the accuracy, flexibility, and speed in providing the most appropriate decision making-strategy for grain growers under field-prone conditions.
Grains Research and Development Corporation (UA00162); Waite Research Institute.
The authors would like to thank Dr. Trung Nguyen and Mr. Song Nguyen from the University of Adelaide for their technical assistance. Additionally, the authors would like to thank Mr. Talal Khan and Dr. Shaghik Atakaramians from the UNSW THz Photonics Group, Sydney for providing insightful spectroscopic information.
The authors declare no conflicts of interest.
1. FAO, “Food and Agriculture Organization of the United Nations. World Agricultural Information Centre, Agriculture data,” (2018).
2. G. A. Slafer, E. H. Satorre, and F. H. Andrade, “Increases in grain yield in bread wheat from breeding and associated physiological changes,” In: G.A. Slafer, ed., Genetic Improvement of Field Crops. (Marcel Dekker, New York, 1994), pp. 1–68.
3. R. Savin and J. L. Molina-Cano, “Changes in malting quality and its determinants in response to abiotic stresses,” In: Barley Science: Recent Advances from Molecular Biology to Agronomy of Yield and Quality. G.A. Slafer, J.L. Molina-Cano, R. Savin, J.L. Araus, and I. Romagosa, eds., (Food Product Press, New York, USA, 2002), pp. 523–550.
4. J. L. Araus, G. A. Slafer, C. Royo, and M. D. Serret, “Breeding for yield potential and stress adaptation in cereals,” Critical Rev. Plant Sci. 27(6), 377–412 (2008). [CrossRef]
5. A. Ferrante, J. Cartelle, R. Savin, and G. A. Slafer, “Yield determination, interplay between major components and yield stability in a traditional and a contemporary wheat across a wide range of environments,” Field Crops Research 203, 114–127 (2017). [CrossRef]
6. R. A. Fischer, “Understanding the physiological basis of yield potential in wheat,” J. Agric. Sci. 145(02), 99–113 (2007). [CrossRef]
7. M. Reynolds, M. J. Foulkes, G. A. Slafer, P. Berry, M. a, J. Parry, J. W. Snape, and W. J. Angus, “Raising yield potential in wheat,” J. Exp. Bot. 60(7), 1899–1918 (2009). [CrossRef]
8. V. O. Sadras, C. Lawson, P. Hooper, and G. K. McDonald, “Contribution of summer rainfall and nitrogen to the yield and water use efficiency of wheat in Mediterranean-type environments of South Australia,” Eur. J. Agron. 36(1), 41–54 (2012). [CrossRef]
9. G. A. Slafer, R. Savin, and V. O. Sadras, “Coarse and fine regulation of wheat yield components in response to genotype and environment,” Field Crops Research 157, 71–83 (2014). [CrossRef]
10. G. A. Slafer and R. Savin., “Physiology of yield,” In: R. Goodman, ed., Encyclopaedia of Plant and Crop Science. (Marcel Dekker, Taylor & Francis Group, New York, USA, 2006).
11. P. De Vita, O. L. D. Nicosia, F. Nigro, C. Platani, C. Riefolo, N. Di Fonzo, and L. Cattivelli, “Breeding progress in morpho-physiological, agronomical and qualitative traits of durum wheat cultivars released in Italy during the 20th century,” Eur. J. Agron. 26(1), 39–53 (2007). [CrossRef]
12. A. Ferrante, R. Savin, and G. A. Slafer, “Differences in yield physiology between modern, well adapted durum wheat cultivars grown under contrasting conditions,” Field Crops Res. 136, 52–64 (2012). [CrossRef]
13. F. Giunta, R. Motzo, and G. Pruneddu, “Trends since 1900 in the yield potential of Italian-bred durum wheat cultivars,”,” Eur. J. Agron. 27(1), 12–24 (2007). [CrossRef]
14. S. Arisnabarreta and D. J. Miralles, “Yield Responsiveness in two-and six-rowed barley grown in contrasting nitrogen environments,” J. Agron. Crop. Sci. 192(3), 178–185 (2006). [CrossRef]
15. V. O. Sadras, “Evolutionary aspects of the trade-off between seed size and number in crops,” Field Crops Res. 100(2-3), 125–138 (2007). [CrossRef]
16. V. O. Sadras and G. A. Slafer, “Environmental modulation of yield components in cereals: Heritabilities reveal a hierarchy of phenotypic plasticities,” Field Crops Res. 127, 215–224 (2012). [CrossRef]
17. J. C. Zadoks, T. T. Chang, and C. F. Konzak, “A decimal code for the growth stages of cereals,” Weed Res. 14(6), 415–421 (1974). [CrossRef]
18. R. A. Fischer, “Number of kernels in wheat crops and the influence of solar radiation and temperature,” J. Agric. Sci. 105(2), 447–461 (1985). [CrossRef]
19. S. Arisnabarreta and D. J. Miralles, “Critical period for grain number establishment of near isogenic lines of two-and six-rowed barley,” Field Crops Res. 107(3), 196–202 (2008). [CrossRef]
20. C. M. Cossani, G. A. Slafer, and R. Savin, “Grain yield and resource use efficiency of bread wheat, barley and durum wheat under Mediterranean environments, ” PhD Thesis. http://hdl.handle.net/10803/8355. (2010).
21. J. Foley, “Frost in the Australian region. Melbourne: J.J. Gourley, Government Printer,” (1945).
22. T. M. Frederiks, J. T. Christopher, S. E. H. Fletcher, and A. K. Borrell, “Post head-emergence frost resistance of barley genotypes in the northern grain region of Australia,” Crop Pasture Sci. 62(9), 736–745 (2011). [CrossRef]
23. T. M. Frederiks, J. T. Christopher, G. L. Harvey, M. W. Sutherland, and A. K. Borrell, “Current and emerging screening methods to identify post-head-emergence frost adaptation in wheat and barley,” J. Exp. Bot. 63(15), 5405–5416 (2012). [CrossRef]
24. T. M. Frederiks, J. T. Christopher, M. W. Sutherland, and A. K. Borrell, “Post-head-emergence frost in wheat and barley: defining the problem, assessing the damage, and identifying resistance,” J. Exp. Bot. 66(12), 3487–3498 (2015). [CrossRef]
25. M. Rebbeck, C. Lynch, P. T. Hayman, and V. O. Sadras, “Delving of sandy surfaced soils reduces frost damage in wheat crops,” Aust. J. Agric. Res. 58(2), 105–112 (2007). [CrossRef]
26. S. J. Crimp, B. Zheng, N. Khimashia, D. L. Gobbett, S. Chapman, M. Howden, and N. Nicholls, “Recent changes in southern Australian frost occurrence: Implications for wheat production risk,” Crop Pasture Sci. 67(8), 801–811 (2016). [CrossRef]
27. B. Zheng, S. C. Chapman, J. T. Christopher, T. M. Frederiks, and K. Chenu, “Frost trends and their estimated impact on yield in the Australian wheatbelt,” J. Exp. Bot. 29, 171–172 (2015). [CrossRef]
28. A. Ferrante, R. Savin, and G. A. Slafer, “Floret development of durum wheat in response to nitrogen availability,” J. Exp. Bot. 61(15), 4351–4359 (2010). [CrossRef]
29. J. L. Araus and J. E. Cairns, “Field high-throughput phenotyping: the new crop breeding frontier,” Trends in Plant Sci. 19(1), 52–61 (2014). [CrossRef]
30. E. M. Perry, J. G. Nuttall, A. J. Wallace, and G. J. Fitzgerald, “In-field methods for rapid detection of frost damage in Australian dryland wheat during the reproductive and grain-filling phase,” Crop Pasture Sci. 68(6), 516–526 (2017). [CrossRef]
31. G. J. Fitzgerald, E. M. Perry, K. C. Flower, J. N. Callow, B. Boruff, A. Delahunty, A. Wallace, and J. Nuttall, “Frost damage assessment in wheat using spectral mixture analysis,” Remote Sens. 11(21), 2476 (2019). [CrossRef]
32. J. G. Nuttall, E. M. Perry, A. J. Delahunty, G. J. O’Leary, K. M. Barlow, and A. J. Wallace, “Frost response in wheat and early detection using proximal sensors,” J. Agro. Crop. Sci. 205(2), 220–234 (2019). [CrossRef]
33. M. E. Murphy, B. Boruff, J. N. Callow, and K. C. Flower, “Detecting frost stress in wheat: A controlled environment hyperspectral study on wheat plant components and implications for multispectral field sensing,” Remote Sens. 12(3), 477 (2020). [CrossRef]
34. J. Schmidt, J. Claussen, N. Wörlein, A. Eggert, D. Fleury, T. Garnett, and S. Gerth, “Drought and heat stress tolerance screening in wheat using computed tomography,” Plant Methods 16(1), 15 (2020). [CrossRef]
35. P. U. Jepsen, D. G. Cooke, and M. Koch, “Terahertz spectroscopy and imaging - Modern techniques and applications,” Laser Photonics Rev. 6(3), 418 (2012). [CrossRef]
36. H. Elke, F. Friedhoff, and C. Hirschmann, “Method for classifying seeds,” U.S. Patent No. 9,857,297.,” (2018).
37. J. J. Liu, Z. Li, F. R. Hu, T. Chen, and A. J. Zhu, “A THz spectroscopy nondestructive identification method for transgenic cotton seed based on GA-SVM,” Opt. Quantum Electron. 47(2), 313–322 (2015). [CrossRef]
38. X. H. Hu, W. H. Lang, W. Liu, X. Xu, J. B. Yang, and L. Zheng, “A non-destructive terahertz spectroscopy-based method for transgenic rice seed discrimination via sparse representation,” J. Infrared Milli. Terahz. Waves 38(8), 980–991 (2017). [CrossRef]
39. H. Luo, J. P. Zhu, W. N. Xu, and M. J. Cui, “Identification of soybean varieties by terahertz spectroscopy and integrated learning method,” Optik 184, 177–184 (2019). [CrossRef]
40. X. D. Sun and J. B. Liu, “Measurement of plumpness for intact sunflower seed using terahertz transmittance imaging,” J. Infrared Milli. Terahz. Waves 41(3), 307–321 (2020). [CrossRef]
41. Y. L. Yang, S. L. Zhou, J. Song, J. Huang, G. L. Li, and S. P. Zhu, “Feasibility of terahertz spectroscopy for hybrid purity verification of rice seeds,” Int. J. Agric. Biol. Eng. 11(5), 65–69 (2018). [CrossRef]
42. M. Shastri, P. Marcharla, S. Thigale, E. Goldys, and S. Atakaramians, “Towards identifying grass seed infestation in the Australian sheep using terahertz radiation,” Conference on Laser and Electro-optics Pacific Rim (CLEO-PR), Sydney, Australia (2020).
43. H. Ge, Y. Jiang, Z. Xu, F. Lian, Y. Zhang, and S. Xia, “Identification of wheat quality using THz spectrum,” Opt. Express 22(10), 12533–12544 (2014). [CrossRef]
44. H. Y. Ge, Y. Y. Jiang, and Y. Zhang, “THz spectroscopic investigation of wheat-quality by using multi-source data fusion,” Sensors 18(11), 3945 (2018). [CrossRef]
45. H. S. Chua, J. Obradovic, A. D. Haigh, P. C. Upadhya, O. Hirsch, D. Crawley, A. A. P. Gibson, L. F. Gladden, and E. H. Linfield, “Terahertz time-domain spectroscopy of crushed wheat grain,” in IEEE MTT-S International Microwave Symposium Digest, p. 4 pp (2005).
46. H. Huang, P. Qiu, S. Panezai, S. Hao, D. Zhang, Y. Yang, Y. Ma, H. Gao, L. Gao, Z. J. O. Zhang, and L. Technology, “Continuous-wave terahertz high-resolution imaging via synthetic hologram extrapolation method using pyroelectric detector,” Opt. Laser Technol. 120, 105683 (2019). [CrossRef]
47. M. Wan, H. Yuan, J. J. Healy, and J. T. Sheridan, “Terahertz confocal imaging: Polarization and sectioning characteristics,” Opt. Lasers Eng. 134, 106182 (2020). [CrossRef]
48. R. I. Stantchev, X. Yu, T. Blu, and E. J. N. C. Pickwell-MacPherson, “Real-time terahertz imaging with a single-pixel detector,” Nat. Commun. 11(1), 2535 (2020). [CrossRef]
51. N. A. Cocks, T. J. March, T. B. Biddulph, A. B. Smith, and B. R. Cullis, “The provision of grower and breeder information on the frost susceptibility of wheat in Australia,” J. Agric. Sci. 157(5), 382–398 (2019). [CrossRef]
53. S. R. Waddington, P. M. Cartwright, and P. C. Wall, “A quantitative scale of spike initial and pistil development in barley and wheat,” Ann. Bot. 51(1), 119–130 (1983). [CrossRef]
54. A. Ferrante, R. Savin, and G. A. Slafer, “Floret development and grain setting differences between modern durum wheats under contrasting nitrogen availability,” J. Exp. Bot. 64(1), 169–184 (2013). [CrossRef]
55. A. Ferrante, R. Savin, and G. A. Slafer, “Floret development and spike fertility in wheat: Differences between cultivars of contrasting yield potential and their sensitivity to photoperiod and soil N,” Field Crops Res. 256, 107908 (2020). [CrossRef]
56. Y. Yang, A. Shutler, and D. J. O. E. Grischkowsky, “Measurement of the transmission of the atmosphere from 0.2 to 2 THz,” Opt. Express 19(9), 8830–8838 (2011). [CrossRef]