Aphrophoridae Foam Detection using YOLOv11 for Sustainable Management of Xylella Fastidiosa Vectors

The growing threat of vector-borne pathogens, along with the impacts of pesticides on environmental ecosystems and human health, has significantly increased the interest in sustainable pest control strategies. This paper presents a novel approach for the early detection of Aphrophoridae, generally known as spittlebugs, which are the primary vectors of Xylella fastidiosa (Xf) in Europe. Following the infection, Xf bacteria colonies occlude the wood vessels and prevent water from reaching the canopy, resulting in leaf burning and plant death. The proposed solution exploits recent advances in deep learning techniques to automatically identify the distinctive white foamy secretion (“spittle”) produced by Aphrophoridae nymphs in images acquired under field conditions. Specifically, the base versions of YOLOv11 and YOLOv12 are first compared to establish a baseline to determine the most effective approach for the case study. Then, a variant of YOLOv11 is tested by integrating a Convolutional Block Attention Module (CBAM) into the neck of the network to allow for more focus both on relevant areas and channels of the fused feature maps before the classification head. The image processing system is intended to guide the action and assess the efficiency of an aeroaulic machine able to generate an airstream with shape and thrust appropriate to remove the target organisms for sustainable management of Xf vectors. Experimental tests were performed in an area covered by wild tall grass, showing that, on average, all the considered models provide a precision above 50%, while achieving recall scores ranging between 35 and 45%. However, the use of CBAM attention provides interesting enhancements to the YOLOv11 model, highlighting its potential to be effectively deployed under constrained settings, while retaining relatively high accuracy and F1 score.