Assessment of cotton and sorghum stand establishment using UAV-based multispectral and DSLR-based RGB imagery

Authors: DHAKAL, MADHAV - Mississippi State University; Huang, Yanbo; Locke, Martin; Reddy, Krishna; Moore, Matthew; KRUTZ, JASON - Mississippi State University; GHOLSON, DREW - Mississippi State University; BAJGAIN, RAJEN - Oak Ridge Institute For Science And Education (ORISE)

Submitted to: Agrosystems, Geosciences & Environment
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 1/10/2022
Publication Date: 5/1/2022
Citation: Dhakal, M., Huang, Y., Locke, M.A., Reddy, K.N., Moore, M.T., Krutz, J., Gholson, D., Bajgain, R. 2022. Assessment of cotton and sorghum stand establishment using UAV-based multispectral and DSLR-based RGB imagery. Agrosystems, Geosciences & Environment. 5(2):e20247. https://doi.org/10.1002/agg2.20247.
DOI: https://doi.org/10.1002/agg2.20247

Interpretive Summary: Crop density and canopy cover are the key agronomic traits for row-crops such as cotton and sorghum to assess the stand establishment. The traditional methods of crop density and ground cover assessment are laborious, time-consuming, sometimes impractical, and could be biased. Scientists from Mississippi State University and USDA-ARS have collaboratively developed a machine learning-based method that uses images obtained from Unmanned Aerial Vehicle (UAV) equipped with multispectral sensors and a handheld digital camera to determine cotton and sorghum stand density and canopy cover in early growth stages. The study evaluated the utility of fifteen different vegetation indices (VIs) in broadband and narrowband spectra by comparing the goodness of fit between observed and estimated plant density and canopy cover. The results indicated that excess green minus excess red (ExG-ExR) among broadband VIs and soil line indices [modified soil adjusted VI (MSAVI) & optimized soil adjusted VI (OSAVI)] from narrowband spectra can effectively estimate the plant density and canopy cover within 30 days of planting. With single-lens reflex camera images, an open-source software ImageJ provided a better estimation of crop density and leaf area index (LAI) than the multispectral data analysis. On a small-scale, use of ImageJ-based image analysis would be a rational choice, however, multispectral imagery can provide flexibility in effectively predicting crop attributes with machine learning applications at large-scale production.

Technical Abstract: Plant density and canopy cover are key agronomic traits for cotton (Gossypium hirsutum L.) and sorghum (Sorghum bicolor L. Moench) phenotypic evaluation. The objective of this study was to evaluate utility of broadband red-green-blue (RGB) and narrowband green, red, red-edge, and near-infrared spectral data taken by an unmanned aerial vehicle (UAV), and sRGB taken by a digital single-lens reflex camera for assessing the cotton and sorghum stands. Support Vector Machine was used to analyze UAV images, whereas ImageJ was used for sRGB images. Fifteen vegetation indices (VIs) were evaluated for their accuracy, predictability, and residual yield. All VIs had Cohen’s k > 0.65, F-score > 0.63, and User and Producer accuracy of more than 71 and 69%, respectively. Modified soil-adjusted VI (MSAVI), optimized soil-adjusted VI (OSAVI), chlorophyll index at red-edge (CIRE), and red-edge minus red (RE-R) among narrowband VIs and excess green minus excess red (ExG-ExR) among broadband VIs provided more agreeable estimates of cotton and sorghum density than the remaining VIs with R2 and index of agreement (IoA) up to 0.79 and 0.92, respectively. The estimated canopy cover explained up to 83 and 82% variability in leaf area index (LAI) of cotton and sorghum, respectively. The sRGB image analysis via ImageJ produced R2 from 0.79 to 0.90 and 0.83 to 0.86 and IoA 0.89 to 0.97 and ~0.91 between estimated and observed cotton and sorghum density, respectively. ImageJ analysis for green cover captured up to 82% and 79% variability in cotton and sorghum LAI, respectively. Although ImageJ can give close estimates of crop density and cover, UAV-based narrowband VIs still can provide an agreeable, trouble-free, and time-efficient estimate of these attributes, especially with SAVIs.