Angular effect in proximal sensing of leaf-level chlorophyll content using low-cost DIY visible/near-infrared camera
Introduction
In precision agriculture, plant proximal sensing recently attracts increasing attention, due to its high potentials of supporting a variety of applications on crop monitoring (Barmeier and Schmidhalter, 2016) and plant phenomics (Wang et al., 2017). For this interdisciplinary field, developing more techniques of varying principles efficient for different purposes has become a pressing task. This point has been aware of by the community (Lin, 2015, Rodrigues et al., 2015, Pallottino et al., 2019, Riebe et al., 2019). Among those carried-out studies of fulfilling this task, using visible/near-infrared (VNIR) cameras for accurate measurement of leaf-level chlorophyll content, which serves as an essential indicator of photosynthesis, mutation, stress, and nutrition state of plants (Buddenbaum et al., 2008), is a typical branch of interest.
A few efforts have been made for advancing this branch. Wang et al. (2016) attempted the VNIR sensing technique for making evaluations of water use efficiency in foxtail millets (Setaria italica). Caruso et al. (2017) applied an unmanned aerial vehicle-based VNIR camera for mapping the biophysical and geometrical parameters of grapevine (Sangiovese) canopies. Cesoniene et al. (2019) investigated the potential of VNIR hyperspectral imaging for drawing the chemical and electro-chemical properties of conventionally and organically grown carrots. Bao et al. (2019) integrated the powerful VNIR hyperspectral cameras with many other types of sensing devices into the Enviratron system for plant phenotyping. A variety of vegetation indices (VIs) applicable for taking the advantages of these techniques have been validated as well (Wang et al., 2016, Caruso et al., 2017, Cesoniene et al., 2019, Bao et al., 2019). These contributions have aroused more attention on VNIR-based plant proximal sensing.
However, in practice this branch develops slowly, in subject to the relatively high prices of the related in-market VNIR cameras. In order to overcome this limitation, some researchers have attempted to modify the readily-available RGB (or other modes) cameras of low prices in a do-it-yourself (DIY) way, which, then, display the VNIR-related functions to some extent. Fan et al. (2018) adapted a simple VNIR camera for measuring the variable of leaf area index and then quantifying the growth stages of Italian Ryegrass (Lolium multiflorum). Hong et al. (2019) operated small unmanned aircraft system-based measurements by utilizing a modified camera (containing three new spectral bands, i.e., a near-infrared band (680–780 nm) and two overlapping green and blue bands (400–580 nm)) for detection of the potential drought stress in Creeping Bentgrass (Agrostis stolonifera L.). Such innovative endeavors have expanded the practical applications of various DIY-made VNIR sensors for plant proximal sensing. Further, based on VNIR spectral sensing, Araujo et al. (2013) have proposed a quantitative and quick method for analyzing the effect of applying agricultural lime to soil, and this facilitated better understanding of the crop-growing environments.
In addition to acquisition of low-cost VNIR cameras by modification of low-price RGB ones, enhancement of their proximal sensing performance from other aspects such as system integration and retrieval modelling is also a viable way for propelling the development of this branch. Cogliati et al. (2015) proposed a novel strategy of integrating the automated in-field VNIR spectroscopy systems for continuously and long-term monitoring of canopy reflectance and sun-induced chlorophyll fluorescence. Buddenbaum et al. (2019) explored the so-called “smile effect” (shifts of the wavelength positions across the field of view) in high-resolution VNIR hyperspectral images for better unveiling the possible impacts of herbicides on young Pine (Pinus contorta) trees. Turner et al. (2019) optimized the spectral and spatial resolutions of their used unmanned aerial system-based VNIR imaging sensors for better measurement of Antarctic vegetation. Such improvements proved to be also able to enhance the applicability of VNIR proximal sensing for plant mapping.
However, it is far from enough for these endeavors to play the full roles of such modified VNIR cameras in proximal sensing of various plant properties. To fill this gap, this study was dedicated to exploring their potential on plant measurements, in the case of proximal sensing of leaf-level chlorophyll content based on a low-cost VNIR camera modified in the DIY way. Under this experimental framework, the typically complicated angular effect (Matasci et al., 2015) that is caused by the complexity of leaf inclinations was examined to further reflect the technical potentiality of those DIY-modified VNIR cameras. The holistic aim was to broaden the range of cutting-edge technologies that are efficient for supporting precision agriculture, via exploring and applying the proximal sensing physics and models of optical-biochemical interactions in leaf reflectance.
Section snippets
Camera modification
The low-cost VNIR camera used in this study was generated by modifying an outdated digital RGB camera (PowerShot A490, Canon). As shown in Fig. 1, the original camera was taken apart, and its function of VNIR imaging was achieved by means of DIY modification. The specific procedures comprised dismantling its original infrared-blocking optical filter and then replacing it with a red-light filter (http://www.publiclab.org/wiki/near-infrared-camera). The red-light filter can shield red light and
Leaf images
For the four species, the acquired VNIR images of their sample leaves are individually illustrated in Table 1. It appears that the colors of the leaves and their background plates are different in contrast to the performance of the common imaging modes. The reason is that the color-matching configuration format of these images still followed the original color setting of the used RGB camera. After the operation of image pre-processing, the blue, green, and near-infrared DN values for the pixels
Discussion
The results have validated the feasibility of the tactics of DIY modifying VNIR cameras for retrieving leaf-level chlorophyll content. The modified camera proved to be able to show both better capacity than the common RGB digital cameras that have once been attempted for this task (e.g., Wang et al., 2014) and comparable performance than the actual VNIR cameras (e.g., Turner et al., 2019). Such performance is particularly implicative for satisfying the wide technical demand on crop health
Conclusion
The study has briefly verified both the validity of DIY modifying low-cost RGB cameras as VNIR ones for retrieving leaf-level chlorophyll content and the efficiency of characterizing the traditionally complicated angular effect for further improving the related retrieval models. The derived such modular retrieval models as constituted by the optimal VI sets for different ranges of leaf inclinations, equivalently, leave a referable database of leaf-optical geometries for handling the complex
Declaration of Competing Interest
The authors declare no competing interests.
Acknowledgements
This work was financially supported by the National Natural Science Foundation of China (Grant No. 31870531 and 31670718).
Author Contributions
Y.L. conceived the study. H.W. conducted the analysis. M.J. examined the results. H.W., M.J., Y.L., Y.Y., Y.F., S.L., and Y.L. co-wrote the paper.
Data availability statement
Data will be made available on request.
References (28)
- et al.
Continuous and long-term measurements of reflectance and sun-induced chlorophyll fluorescence by using novel automated field spectroscopy systems
Remote Sens. Environ.
(2015) - et al.
Does remote and proximal optical sensing successfully estimate maize variables? A review
Eur. J. Agron.
(2018) - et al.
A simple visible and near-infrared (V-NIR) camera system for monitoring the leaf area index and growth stage of Italian Ryegrass
Comput. Electron. Agric.
(2018) LiDAR: An important tool for next-generation phenotyping technology of high potential for plant phenomics?
Comput. Electron. Agric.
(2015)- et al.
Understanding angular effects in VHR imagery and their significance for urban land-cover model portability: a study of two multi-angle in-track image sequences
ISPRS J. Photogramm. Remote Sens.
(2015) Twenty five years of remote sensing in precision agriculture: key advances and remaining knowledge gaps
Biosyst. Eng.
(2013)- et al.
Optoelectronic proximal sensing vehicle-mounted technologies in precision agriculture: a review
Comput. Electron. Agric.
(2019) - et al.
Object-based crop identification using multiple vegetation indices, textural features and crop phenology
Remote Sens. Environ.
(2011) - et al.
Proximal soil sensing for precision agriculture: Simultaneous use of electromagnetic induction and gamma radiometrics in contrasting soils
Geoderma
(2015) - et al.
Validation of a low-cost 2d laser scanner in development of a more-affordable mobile terrestrial proximal sensing system for 3D plant structure phenotyping in indoor environment
Comput. Electron. Agric.
(2017)