Improving plant disease detection by harnessing oblique observation angles from multispectral cameras and unmanned aerial vehicles.

• Applicant: Dr. Rene Heim
• Subject Area: Plant Breeding and Plant Pathology
• Term: since 2023
• Project identifier: Deutsche Forschungsgemeinschaft (DFG) - Project number 521313940

Plant diseases affect crop production globally. As a strategy against plant diseases, the system of integrated plant protection has been brought forward. As innovative tools for plant disease detection and monitoring, sensor technologies provide significant potential. Within this context, unmanned aerial vehicles (UAVs) as carrier platforms for optical sensors are already at the center of research. UAVs can be deployed at short time intervals to monitor large areas at an ever-decreasing cost. They are suitable to explore field-scale disease patterns which is practically not feasible for humans. An element overlooked in digital plant disease detection is the use of oblique camera angle observations from UAVs. The current standard is still tilting the camera 90° downwards to collect images from a nadir view. But it is well known that the observation angle has an influence on optical reflectance data especially when collected from plant canopies that are complex 3-dimensional structures. Few studies have shown that UAVs can be used as mobile, lightweight goniometers to provide multi-angular observation datasets. Such data could improve the estimation of yield, leaf area index, and chlorophyll content in potatoes and corn. In digital plant pathology, UAVs have not yet been used to collect and study multi-angular observation data. But doing so would allow us to evaluate a pathosystem (host + pathogen + environment) from multiple sides and not just from straight above. Especially for dicotyl cultivars, such as sugar beet, where leaves grow more upright, oblique observation angles could allow improving current disease detection models as the whole leaf surface would be visible, instead of just sections of it when only nadir view data is collected. Therefore, the proposed project aims to collect nadir and oblique UAV imagery and study the resulting reflectance patterns at the field scale. We will collect multi-angular, multispectral aerial imagery across two growing seasons. Different cultivars of sugar beet will be inoculated with Cercospora beticola, causing Cercospora Leaf Spot, and compared to the same cultivars that will be kept free from disease. The resulting reflectance data will span a range of observation angles. The observation angles will be binned in balanced classes (e.g., 0-10°, 10-20°, …) and used to train classification models to improve disease severity and disease incidence models. Thus, we will be able to reveal optimal observation angles for the classification of diseased and healthy sugar beet plants at different time points during disease development and obtain a deeper understanding of optical properties of pathosystems at field-scale. By selecting multiple cultivars with differing plant geometry and leaf angles, we aim at exploring optimal observation angles for the same disease but for different plant architectures. This knowledge gain will impact many applications in plant disease management and breeding trials.

• DFG Programme: Research Grants