Listening in on insects leads to identification

Sensor uses artificial intelligence and a sound detector based on a curtain of light to monitor bugs in the field

A California-based startup has developed an insect monitoring system that classifies insects by the sound they make.

“When the bugs arrive, we sense them with our unique sensor, and we can tell right away the sex and species of that insect,” said Eamonn Keogh of FarmSense.

“That information is pushed to the cloud and you can access it on your laptop or cellphone and you can slice and dice the information anyway you want, by sex, by species, location in your field, by time of arrival, date of arrival, and so forth.”

The sounds insects make when in flight vary considerably between species, but it is difficult to use microphones in field-based sensors because of environmental noise, primarily wind, Keogh said.

So FarmSense developed a sensor that uses a curtain of light at the opening of the trap and when a bug flies through the light and causes a very specific disruption pattern.

“We have a kind of microphone but it actually records sound bits with light, not the actual sound. We call it pseudo acoustic,” Keogh said.

The sensor tracks the movement of an insect’s wing beats with a laser and a phototransistor array, and then converts the disruption to a sound file, which FarmSense then processes with algorithms capable of identifying the bug associated with the sound.

Keogh said the machine-learning based algorithms are able to detect more information than just how fast the wings are beating to help differentiate specific bugs, but these algorithms had to be trained to do this.

He said for the last few years, the company built a large archive of insect data for the machine-learning algorithms to work off.

“For the last several years we’ve been taking insect larva, put them in a cage with our sensor, and let them hatch. We watched them from birth to death, 24 hours a day under different temperature and lighting conditions, air pressure, hungry versus not hungry and so forth,” Keogh said.

He said other automated in-field insect sensors use sticky traps and cameras that send in images a few times a day, which then need to be counted. But these systems don’t tell you when the bug arrived, and the traps are ineffective when they get saturated full of dead bugs.

“No matter how many we get we can keep on counting them,” Keogh said.

“They are taking photos of dead insects and sometimes they really can’t figure out what insect it is. It’s just some bug. In contrast we’re recording live, dynamic, vigorous flying insects so we get a much richer amount of information.”

The number of sensors needed on a field varies greatly, but one sensor for every four acres to one sensor for every 20 acres is a typical ratio for high value crops.

He said there are 400 to 500 bugs around the world that are harmful to crops, and only about 40 bugs that inflict the most crop damage.

So far, FarmSense has models built for about 20 of these pests, but its research teams are expanding the database and within 10 years want to have sensors suited for every crop in every area of the world.

Keogh said the more information you can put into economic threshold models, including the species, sex, and time the pest arrived on the field, the more useful the modelling will be.

“Many of those models that have inputs for insects, they tend to assume a huge uncertainty in the information and they assume a huge time lag. Those models can be adjusted to be even better when you have less uncertainty in the insect counts, and less of a time leg,” Keogh said.

In the short term, FarmSense will focus on high-value crops such as fruits and nuts, but the company is conducting trials in broad-acre crops.

The company is also exploring how the sensor can be used to monitor pollination.

“Pollination for a service is a huge thing here in California,” Keogh said.

“If you’re a grower and you’re paying a pollination service to pollinate your almonds, you want to make sure you have enough beneficial (insects), the right ratio and so forth.”

The FarmSense insect trap uses solar power and cellular connectivity, so once it’s installed it doesn’t need to be visited again throughout the growing season.

In areas with poor connectivity Low Power, Wide Area (LPWA) technology can be used.

“We do all the processing, the computations, directly on the trap and we only transmit the insect counts. What that means is we are only sending a few kilobytes, like a text message from these traps, not heavy-duty audio or video that would be expensive to do,” Keogh said.

Users can set the trap to upload the insect counts as often as they please, from instantaneous upload to once per day.

As the company updates their models, it remotely uploads the information to the sensors already installed.

The bug sensors also have temperature and humidity sensors that users can access.

“This info is valuable to use for the bugs, but it’s also valuable to the grower to help them figure out the microclimates in their fields,” Keogh said.

The FarmSense insect sensor will be available to purchase online in 2021, and it will be set up for a few dozen pests.

About the author


Stories from our other publications