Estimation of Cotton and Sorghum Crop Density and Cover at Early Vegetative Stages Using Unmanned Aerial Vehicle Imagery

Abstract:

Crop density and canopy cover are key agronomic traits for cotton (Gossypium hirsutum L.) and sorghum (Sorghum bicolor L.) evaluation and water and nutrient management. The traditional method of assessment of these parameters is labor-intensive and time-consuming. An Unmanned Aerial Vehicle (UAV) equipped with multispectral sensors can precisely and effectively determine crop stand density and cover in early growth stages. The objective of this study is to develop a machine vision-based method to automate the survey of cotton and sorghum density and canopy cover at early growth stages. Multispectral images were taken with a high-resolution camera mounted on a UAV at broadband RGB and narrowband green, red, red edge, and near-infrared region. Images were acquired multiple times in 2019 and 2020 from the USDA-ARS long-term agroecosystem research (LTAR) project, Stoneville, MS that was designed to investigate the effects of cover crop, tillage, and crop rotation on cotton and/or sorghum. Spatial analysis was performed for each image using ArcGIS. Sorghum and cotton plant pixels were extracted from background (bare soil) using NDVI, Excess Green Index, and Otsu thresholding methods. Morphological features were calculated from images using a shape features in ArcGIS to estimate plant density, canopy cover, and canopy height. Estimates will be validated with field data from a total of 165 crop plots from two experiments, which include manually determined stand count, canopy cover estimated using ImageJ software taken from an RGB camera, and leaf area determined using a Leaf Area Meter. Initial results (with 48 images from 2019) suggested that UAV imagery can be used to accurately estimate crop canopy cover with r² = 0.86 and 0.80 and root mean square error (RMSE) = 9.4 and 8.1% for sorghum and cotton, respectively. Crop density estimates strongly correlated with field measurements (R² = 0.78 and 0.85, RMSE = 0.6 and 1.0 plants m^–2 for cotton and sorghum, respectively), whereas leaf area index estimates had a weaker relationship (R² = 0.53 and 0.44, RMSE = 0.11 and 0.18 m² m^–2 for cotton and sorghum, respectively). The UAV image analysis method is a promising tool for use as an effective phenotyping method to assess cotton and sorghum crop development.