Student Spotlight on Yanqiu Yang:  Creating a Smart Trap: Using artificial intelligence to track an invasive species  

Can technology slow the spread of invasive species?  

Yanqiu Yang, a PhD candidate in Agricultural and Biological Engineering, is working to answer this question. To slow the spread, it is important to track populations. Then, actions can be taken to contain the spread. However, tracking invasive populations is not an easy task. Invasive species are often difficult to identify and require experts for a confident identification. This is a time- and labor-consuming method that results in a low number of identifications. Relying on humans to identify specimens reduces how quickly actions can be taken against invasive species.  

To solve this, Yanqiu is creating an artificial intelligence (AI) algorithm to identify an invasive moth species called the box tree moth. Then, she will incorporate this algorithm into two complementary systems. First, she will create a web portal that uses AI to identify the box tree moth from a user-submitted image. Second, she will design a trap that attracts the moth and then uses AI to identify what species is in the trap.  

“By integrating AI with camera traps and our web portal, we can speed up detection and reduce misidentifications,” Yanqiu explains, “This collective effort will enhance our ability to monitor and control the spread of invasive species more effectively.” 

Box tree moths are an invasive species that damage boxwood shrubs, a common ornamental plant. Box tree moths are particularly difficult to identify because they look very similar to native moths, like the melonworm moth. Identification relies on collecting moths in traps, sending moths to entomologists, and these entomologists sifting through samples of collected insects to determine if any are the box tree moth. This method is time-consuming, and box tree moths are likely not identified until there is already a sizable population in an area. Yanqiu’s overall goal is to design a system that helps box tree growers and others more effectively monitor the moth population across the eastern United States. 

Many invasive species have the potential to be monitored using similar technology. Yanqiu’s work to create AI monitoring algorithms will make it easier to apply this technology to other species.  

 The first step of Yanqiu’s project — creating a web portal — is currently available for use. MothNET is a portal where AI will analyze and identify the species present in an uploaded picture. This technology is free to anyone who wants to use it: gardeners, nursery owners, or even general moth enthusiasts.  

One challenge with creating an AI algorithm is that the algorithm needs to be “trained” so that it can accurately identify the species. This process takes many hours of work and hundreds to thousands of training images. When MothNET makes an identification, it provides a measure of confidence about that identification. The measure of confidence shows how likely it is that MothNET’s identification is correct. For example, 90% confidence means that the identification is most likely accurate. On the other hand, a score of 25% means that there is a high chance the identification is not accurate. In the future, Yanqiu plans to expand this portal to identify more moth species and to continue training the algorithm to be more accurate.  

Yanqiu’s trap will work by using commercially available pheromones to lure box tree moths into the trap, which are species-specific chemicals that the moths use to communicate with each other. When a moth flies into the trap, a camera will take a picture. Then, the computer attached to the camera will use the AI algorithm to identify the moth. The trap will even be wireless, so users can see the identifications without emptying the trap.  While other traps already exist to lure and capture box tree moths, this will be among the first to incorporate such sophisticated identification technology.

The camera traps will be deployed throughout the eastern United States to track moth populations. Once box tree moths are confirmed in an area, quarantine zones can be put in place. In a quarantine zone, the shipment of boxwoods is limited to decrease the chance of box tree moths spreading from that area.  

Creating these systems is a complicated endeavor. Luckily, Yanqiu is uniquely suited for this task. She has a background in both mechanical engineering and digital image processing, where she developed the skills that she will need. She will partner with USDA-APHIS (Animal and Plant Health Inspection Service) labs to design these traps. This partnership will give her insight on invasive species management at a federal level. These experiences and partnerships will help Yanqiu to design technology that can confidently identify box tree moths.  

In the future, this type of technology can be applied to controlling and tracking populations of other invasive species. In the meantime, if you want to take part in tracking box tree moths, give MothNET a try! Take a picture, upload it to this link (https://diagnose-invasive.vmhost.psu.edu/), and see what MothNET thinks.  

___________

Yanqiu Yang is a Fellow in the INSECT NET program and a PhD student mentored by Professor Paul Heinemann in the Department of Agricultural and Biological Engineering at Penn State.  Her research is also supported by funding from the Technology for Living Systems Center at Penn State. 

This article was written by Sophia Mucciolo, a fellow in the INSECT NET program and a PhD student mentored by Sara Hermann in the Ecology Graduate Program and Department of Entomology. It was prepared as part of the INSECT NET Science Communication workshop series coordinated by Drs. Christina Grozinger and Natalie Boyle in summer 2024.