S14 - Session P4 - Providing insect activity forecast to growers using automated traps network, deep learning and neural networks

Authors: Vincent Joly *, Mateja Stefancic, Matej Stefancic

Insect monitoring methods have evolved considerably in the recent years, driven by development of integrated pest management methods. Insect are monitored with pheromone-based traps (when available). This requires time consuming traps checking on a weekly or bi-weekly basis, along with manual compilation and analysis of captures information. New biological and soft chemical crop protection products are increasingly used, driven by stricter food safety regulations. Among consequences of this evolution, time window for application is getting limited to target pest at specific development stage. If done out of optimal timing, it reduces dramatically its efficacy and increases the risk of stunning into pest resistance. We demonstrate how a novel concept of placing networks of automated traps, collecting data and processing them with advanced deep learning and neural networks algorithms significantly improve pest population understanding. This enabling faster conversion to more environment friendly crop protection. Consistent trap network in an area gives a better view on pest pressure situation and indicates trends of pest occurrence in the region. In 2022, many Trapview networks for various pests are active around the world, among them on following pest/crops combinations : H. armigera and T. absoluta on tomato / A. gamma , Spodoptera sp. and H. armigera in lettuce / L. botrana on grapes / C. pomonella on apples…