FarmSense Machine Learning
Gnawing, burrowing, infecting: The damages caused to agriculture by insect pests like the Japanese beetle (pictured above) exceed $100 billion every year, according to the Agricultural Research Service of the USDA. And along with plant diseases, which the exoskeleton buggers can also transmit, arthropods account for the annual 40% loss of agricultural production worldwide.
Enter FarmSense, a Riverside, California-based agtech startup attempting to solve the insect pest problem. The company creates optical sensors and novel classification systems based on machine learning algorithms to identify and track insects in real-time. The key here: real-time information.
They claim real-time information provided by their sensors allows for early detection and thus the timely deployment of pest-management tools, such as insecticide or biocontrols. The current mechanical traps used for monitoring may only yield important intel 10 to 14 days after the bugs’ arrival.
“Some of these bugs only live as adults for like five days, so by the time you know you have a problem, the problem has already taken root and is now a bigger problem,”
Eamonn Keogh, a co-founder of FarmSense.
How they can provide the information critical for achieving those better outcomes is a bit complicated.
Currently being tested and researched in almond orchards in Southern California thanks to a Small Business Innovation Research grant, their newest sensor, termed the FlightSensor, is best understood when considering where Keogh got the idea for it: James Bond and Cold War espionage.
Keogh explained how Russian spies would use lasers, poised on glass window panes, to pick up on vibrations caused by people’s voices. Then a sensor would translate that information, providing rough intel on what was going on in the room.
However, instead of reading vibrations, the FlightSensor uses light curtains and shadows within a small tunnel that the insects are drawn into by attractants. On one side of the sensor is a light source and on the other is the optical sensor. The sensor measures how much light is occluded, or rather how much makes it across when an insect flies inside. That data is turned into audio and analyzed by machine learning algorithms in the cloud.
According to FarmSense, the sensor, which is designed to look like old analog devices for ease-of-use by growers, does not pick up on ambient noises, such as wind or rainfall.
FarmSense Machine Learning
