Uncrewed Drones, Backed by YOLO, Take to the Skies to Hunt Down the Pine Processionary Moth

Researchers from the Universities of Lisbon, Bordeaux, and Padova, and Telespazio France have demonstrated how uncrewed aerial drones can help fight back against the pine processionary moth — thanks to deep learning.

"Early detection of insect infestation is a key to the adoption of control measures appropriated to each local condition," the researchers explain by way of background to the project, before highlighting the problem with the otherwise tempting approach of using remote sensing rather than boots on the ground. "The use of remote sensing was recommended for a quick scanning of large areas, although it does not work well with signals bearing low intensity or items that are difficult to detect."

That's where drones come in. The team set about deploying uncrewed aerial vehicles (UAVs) equipped with camera sensors to record footage of forests in France, Italy, and Portugal known to be targets for the pine processionary moth — a pest species the larvae of which releases stinging hairs, representing a risk to public health.

Thankfully, the pine processionary moth has another rather more useful behavior: it spins conspicuous silk nests on pine trees at the beginning of winter, easily visible to the naked eye. Replacing said eye and its connected brain with UAVs and a deep learning system based on the You Only Look Once (YOLO) algorithm to spot the nests from the air far more efficiently and over a greater area than has previously been possible.

The researchers tested two methods of analyzing the footage captured by the drones — region-based convolutional neural networks (R-CNNs) and YOLO — before settling on their final approach. This, testing showed, delivered F1-scores of 0.826 for detecting the presence of nests and 0.696 for detecting their absence, more than good enough to be valuable for early intervention should an infestation rear its head.

There is, however, one small catch: nests built under a tree's canopy may not be visible from the air, though the reverse is true for those on the ground looking for nests built higher up. "The detection of all the nests that can be present on a tree is not achievable with either UAV scanning or traditional ground observation," the team explains, "therefore the integration of the methods may allow the complete efficiency of the surveillance."

The team's work has been published under open access terms in the journal NeoBiota.