Integrated invasive pest detection using drones and machine learning

Agroecosystems in the United States face major threats by invasive pests, including insects, weeds, and plant diseases. Early detection and rapid response (EDRR) has been widely used to detect, contain, eradicate, and manage invasive pests before they become widely established. However, efficient detection and management of the most invasive pests is still challenging, mainly because current EDRR relies on limited remote sensing with satellites and ground survey. Here, we present a new "Integrated Invasive Pest Detection System” by incorporating remote sensing, conventional ground surveys, and artificial intelligence (AI), which can obtain and process big data at multiple spatial and temporal scales to help invasive pest detection. The goal of this study was to develop a new system to efficiently detect various invasive pests in agroecosystems using EDRR empowered by aerospace technology and machine learning. In this study, we integrated, unmanned aerial system (UAS, drone), and conventional ground survey into a system for invasive pest detection. A 2-year field study was conducted to detect signs and damages of deciduous trees by insect pests including sponge moth and emerald ash borer. The results of this study indicated that aerial images obtained from satellites and UAS could detect various levels of damages by the insects, and machine learning could help automated detection of invasive pests. Two major outcomes of this study were (1) the Integrated Invasive Pest Survey system for large-scale detection of invasive pests and (2) UAS/AI-empowered EDRR tool for decision making on invasive pest survey and management.