Development of a spraying drone that can detect pine processionary moth nest by deep learning

Pine processionary moth (Thaumetopoea pityocampa and Thaumetopoea wilkinsoni) is one of the important pest species threatening forests especially in the Middle East, Central and Southern Europe and North Africa. This pest, which is also found in the Mediterranean, Aegean, Marmara and Black Sea regions of Turkey, feeds on the leaves of coniferous trees. They slow down the growth of trees they infest in terms of diameter, height and volume. In addition, these pests can be harmful for humans and other mammals due to their allergenic effect. Considering that more than half of forests in our country are composed of coniferous species, the size of pine processionary moth threat is more clearly understood. In this context, a combat tool is developed that aims to detect and spray the sacs containing pine processionary beetle larvae from the air by a drone. In this way, ecological and material damages that may occur by spraying entire tree are minimized by spraying only the target area. In the study, there is a camera placed on specially designed drone to take images and a liquid spray system that will perform spraying function. Deep learning technique is used to detect sacs on the images. By using YOLOv7 network, training, and testing of the previously created dataset are carried out. The success metric of mAP@0.5 in finding pine sacs is determined as 92.5%. A fungus species in biopesticide category is used for the destruction of pine processionary beetle larvae in the detected pine nests.