Abstract
Soybean (Glycine max (L.) Merr.) is a leguminous and oil crop with rapidly growing importance and demand in Africa following the increasing demand for oil and livestock and poultry feed in sub-Saharan Africa. However, soybean productivity is low in most countries of sub-Saharan Africa, especially in West Africa, where productivity is below one ton per ha. Hence, concerted soybean varietal development and testing efforts have been underway by the International Institute of Tropical Agriculture (IITA), collaborating with the various African and US-based soybean breeding programs. Integrating new varietal evaluation approaches based on advanced phenotyping techniques into IITA's soybean breeding program is crucial for designing efficient crop genetic improvement techniques. Hence, this work aims to investigate machine learning (ML) models and Unmanned Aerial vehicles (UAVs) to aid rapid high throughput phenotypic workflow for soybean yield estimation. We acquired multispectral images through a Sequoia® camera aboard a senseFly eBee X UAV from five variety trials during the 2020 growing season in Nigeria. UAV-based spectral bands, canopy height, vegetation indices (VI), and texture features were generated by gray level co-occurrence matrix (GLCM) and integrated to predict crop grain yield using five machine learning (ML) regression models, including Cubist, Extreme Gradient Boosting (XGBoost), Stochastic Gradient Boosting (GBM), Support vector machine (SVM), and Random Forest (RF). The main findings are the textural information generated using gray level co-occurrence matrix (GLCM) slightly outperformed predictors based mainly on vegetation indices (VI) and provided a promising alternative to the conventional use of VI in crop yield estimation. All the five ML models performed moderately well in predicting grain yield for all the soybean trials investigated, though the Cubist and RF model stood out, with R2 reaching 0.89. The study provides a framework to perform crop breeding trial assessments more effectively and consistently at high spatial scales that African crop breeding programs did not commonly apply. The workflow can also be successfully modified and applied for high throughput phenotyping of breeding platforms in other crops.
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
More From: Remote Sensing Applications: Society and Environment
Disclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.