Artificial intelligence may help reduce the time and effort required to effectively monitor the brown marmorated stink bug (Halyomorpha halys) in orchards, according to pan-European research — but work is needed to perfect approaches.
The bug, which originates from Asia, is considered one of the biggest invasive threats to European growers due its broad diet and potential for expanding into new territories. Current control methods mainly rest on monitoring using trapping techniques and the use of conventional pesticides.
In study findings presented at 5th International Workshop on Intelligent Systems for the Internet of Things, scientists reported that combining conventional pheromone trap monitoring with a camera and image identification based on machine learning provided “very satisfactory” positive identification of H. halys.
However, so far, identification of the pest species using drone-mounted cameras flying over trees has not provided suitable levels of accuracy. This was the first time identification of the bug has ever been attempted using drones, and continued research, as part of the EU-funded HALY.ID project, will attempt to overcome initial issues by developing more sophisticated machine learning algorithms.
Aiming for intelligent i.d.
The use of the technologies in the work is aimed at reducing the significant amount of time and labour involved in current monitoring approaches.
Drones can cover more ground than human observers, they noted, and are faster and more able to navigate terrains than ground robots. For identifying specimens attracted by pheromones and caught in sticky traps, cameras linked to AI can theoretically overcome the need for high levels of entomological skill in monitoring work.
The scientists dedicated considerable effort in the project so far to selecting the most appropriate cameras to use in the various technologies, in order to identify the key features of H. halys. Then, in order to effectively ‘train’ the deep learning algorithm, they used images of both H. halys and Nezara viridula, another common stink bug in Italy, where the pest has had a large impact.
Additional work by the team aims to optimise sensors to monitor the microclimate of orchards, as a means of improving predictions of the bug’s presence.
