Artificial intelligence is still in its infancy in agriculture, but the technology is being used through such tools as drones, sensors and hyperspectral cameras.

Speakers at a forum on artificial intelligence in agriculture held April 26 at the North Carolina Biotechnology Center in Research Triangle Park see great potential for both artificial intelligence and machine learning to increase yields, fight pests and battle drought. They say the technology is needed because 60 percent more food will be needed to feed a world population that is expected to reach 10 billion people by 2050.

“We cannot increase the amount of land; that is fixed. What we can do is try to improve per hectare yield by using less water while still increasing yields. We are getting there by using a lot of this technology,” said Ranga Raju Vatsavai, associate professor of computer science at North Carolina State University.

Vatsavai said technology is in the works to use artificial intelligence to schedule irrigation to both conserve water and provide moisture to plants at the right place and right time. “Right now we are not there, but we are getting there,” he said.

Huge advances in computer technology and computing power is already benefiting agriculture, he said. For example, Vatsavai is using a super computer the size of credit card attached to a drone that provides on-demand, near real-time event monitoring. The credit-card sized computer has a staggering teraflop of computing power yet only requires 15 watts of power consumption.

“I can put this small super computing device on board a small drone and as the drone is flying, I can analyze the data and identify if there is any stress in the field,” Vatsavai said.

Solmaz Hajmohammadi, image analytics, algorithm/software developer with LemnaTec in Research Triangle Park, N.C., says both artificial intelligence and machine learning can be used to increase yields. LenmaTec is utilizing phenotyping data gathered from the field to find ways to boost yields.

The company uses a hyperspectral camera to gather data from an entire field that provides color coded images that can differentiate plant parts and differentiate plants from non-plants. The camera box is connected with sensors and can measure the water content and chlorophyll content of plants.

“It provides a lot detailed information and shows how the plant is growing,” she explained.

Hajmohammadi said machine learning could be used to classify different types of fruit, detect diseases and create vegetation indices.

Carson Roberts, senior applications engineer for Headwall Photonics in Boston, said hyperspectral imaging is valuable because it can provide spectral analysis of different characteristics of plants, based on color. The technology is already being used in almond processing and apple processing.

In apple production, hyperspectral imaging is used to detect bruises by visible reflect ants while fecal contamination to the fruit is measure by fluorescents. Roberts said this is valuable because if you are squeezing apples for juice, you don’t want any E Coli in the juice. Headwall Photonics is doing its apple work in collaboration with USDA’s Agricultural Research Service.

Spectral analysis is also being used to sort almonds. The analysis provides a pixel by pixel analysis where almonds show up as the color blue in the spectral imager. “The data from the spectral imager is sent to a robot with a vacuum tool that sucks out everything that’s not almonds,” he explained.