Early detection of Cercospora beticola and powdery mildew diseases in sugar beet using uncrewed aerial vehicle-based remote sensing and machine learning

Abstract

Background Agricultural production is crucial for nutrition, but it frequently faces challenges such as decreased yield, quality, and overall output due to the adverse effects of diseases and pests. Remote sensing technologies have emerged as valuable tools for diagnosing and monitoring these issues. They offer significant advantages over traditional methods, which are often time-consuming and limited in sampling. High-resolution images from drones and satellites provide fast and accurate solutions for detecting and diagnosing crops' health and identifying pests and diseases affecting them. Methods The research focused on the early detection of Cercospora leaf spot (Cercospora beticola Sacc.) and powdery mildew (Erysiphe betae (Va & ncaron;ha) Weltzien), which cause significant economic losses in sugar beet before visible symptoms emerge. The study was accomplished by capturing images of uncrewed aerial vehicle (UAV) in field conditions. To effectively evaluate different detection methods in agricultural contexts, the study targeted two key areas: (1) monitoring Cercospora in fields without pesticide application, utilizing the Metos climate station early warning system alongside UAV-based image analysis, and (2) monitoring powdery mildew, which involved visual disease detection and targeted spraying based on UAV image processing. Trial plots were established for this purpose, with six replications for each method. Results UAV-based images show that Normalized Difference Vegetation Index values in leaves decreased before disease onset. This change is an important warning sign for the emergence of the disease. Additionally, the study demonstrated that early detection of diseases is possible using K-nearest neighbors and logistic regression algorithms, exhibiting high discrimination and predictive accuracy.