Fall armyworm moth identification using light-trap cameras and deep learning

Flying insect pests are among the most destructive threats to Australian agriculture, causing considerable crop and financial losses each year. Effective control of insects requires early detection and intervention before infestations have time to get out of hand.

For the Summer of 2024-25, we have a summer scholar, Bill Scott, who is working on a monitoring system which uses UV light to attract a wide range of flying insects and a custom-trained AI model to detect, identify, and count important insect pests.  Bill is exploring the potential to train the AI model on images of lab-raised moth pests including the cotton bollworm (Helicoverpa armigera), tobacco cutwom (Spodoptera litura) and fall armyworm (Spodoptera frugiperda) and testing the model’s ability to differentiate between similar looking species of moth both in the lab and in real-world contexts.

 This technology has applications in biosecurity and pest monitoring. Of particular interest is the North American fall armyworm (FAW) moth, which arrived in 2020 and has caused widespread losses in cereal crops, particularly in Northern Australia, and continues to expand its range throughout the country. We hope to improve the detection of movements of mobile pests like the FAW into and within Australia and better understand how they spread and what factors affect their population numbers. This is one of many approaches being trialled to detect and monitor pest movement.

 

 

Project Title: Fall Armyworm Identification Using Light-Trap Cameras and Deep Learning
Project Date Completion: February 2025
Project Supervisor: Mubin Ul Haque
Project Co-Supervisors and Technical Assistance: Jessa Thurman, Chathurika Amarithunga, and Andy Wang