Writer: Justin Bannister, 575-646-5981, email@example.com
Digging through a bucket of bugs isn't a glamorous job, but the U.S. Department of Agriculture pays someone to do just that. The good news is new technology at New Mexico State University could soon turn this job over to a machine.
The USDA takes samples of insects from forests across the country in large bug traps. Each has thousands of insects inside, and it takes a professional an entire week to sort though it. Every bug in the sample must be classified and counted.
Now, a computer-vision-guided robotic arm at NMSU is able to do this same work, but much faster.
"It takes a person a week to go through those traps; this arm can do that same work in about an hour," said Jeff Drake, an engineer for the USDA and an affiliated university faculty member. "This will save lots of time and lots of money when looking for pests."
The arm is still under development, but can already recognize 30 different insect genera, and most importantly, Drake's team is working to identify the 10 most common species of bark beetle. "Those 10 species account for 95 percent of all bugs found inside the USDA's traps," Drake said.
Bark beetles are found in almost any forest. They bore into the outer bark of trees to feed off the inner bark and lay their eggs. Experts say they can actually help some forests by killing off weaker trees, helping decompose wood and making room for new, healthier trees.
The problem comes when new, invasive species of bark beetles arrive in an area. These new species often won't have natural predators, so their numbers can multiply quickly. Often, the bugs can overwhelm even healthy trees and wipe out large sections of forests.
"The idea to do this really showed up before the technology," said Drake. He said researchers at NMSU's Biological Control Laboratory have long been looking for a way to use computer vision to sort and classify objects. Now Drake has written a computer program using several digital cameras and a mechanical arm to do just that.
Drake's robotic arm uses two digital video cameras mounted above and below a glass tray to "see" what is in front of it. The camera's images are then sent to a computer that recognizes unique features and tells the arm how to sort them. A small suction cup on the end of the arm lets it pick up and drop the insects into individual cups.
The arm is still not 100 percent accurate, but Drake is working on it. If the arm ever finds something it doesn't recognize, it can pick it up, put it under a microscope, and show it to a taxonomist anywhere in the country via the Internet for identification. Once the arm is told what kind of insect it has, it is then placed into the proper category.
In the future, Drake plans to add more cameras, measuring infrared and ultraviolet light, to give the arm more functionality. The extra "eyes" would let the computer get a better look at insects and make it easier to identify them.
Once the arm is perfected, Drake envisions more than a dozen being built and set up in central locations around the country. Then, the USDA and other agencies can send their samples to technicians at those locations. From there, the arms will count samples, sort them, and the results can be sent back to the researchers. The whole process is streamlined and bug-counting humans are free to do other, less icky jobs.
© 2013 New Mexico State University Board of Regents
NMSU - All About Discovery!