APPLICATION OF AI AND UAV IMAGING FOR ASSESSING INFESTATION LEVELS OF Cameraria ohridella IN URBAN ENVIRONMENTS

Abstract

The horse chestnut leaf miner (Cameraria ohridella Deschka et Dimic) is a major pest affecting Aesculus hippocastanum L. in urban areas across Europe. To design effective management strategies, it is essential to accurately assess the level of infestation. However, current observations rely on visual assessments which are subjective, timeconsuming, and not suitable for large-scale monitoring. The research explores the use of AI and UAVs for assessing infestation levels of horse chestnut trees with C. ohridella in urban environments. The infestation severity is classified using convolutional neural networks (CNNs) which are trained with a standardized scale of damage assessment on leaf images, scanned leaf images, and RGB and thermal images from UAVs. The leaf-based assessment and crown thermal indices are compared sufficiently that the approach can assist in early detection of infestation and scalable monitoring. The findings indicate that there exists a significant correlation between the AI predictions and those made by the expert assessments. The thermal anomalies used in the predictive model also show a good relationship with infestation levels. For monitoring C. ohridella infestations, the approach offers a scalable, reproducible, and cost-effective solution. There is a strong possibility of this synergy being included in pest control in urban green infrastructure management and early warning systems and assessment of ecosystem services provided by urban horse chestnut trees.