Scientists from the Skoltech Middle for Computational and Knowledge-Intensive Science and Engineering (CDISE) and the Skoltech Digital Agriculture Laboratory and their collaborators from the German Aerospace Middle (DLR) have developed a synthetic intelligence (AI) system that allows processing photographs from autonomous greenhouses, monitoring plant progress and automating the cultivation course of. Their analysis was printed within the journal IEEE Sensors.
Fashionable know-how has lengthy change into a fixture in all spheres of human life on Earth. Reaching out to different planets is a brand new problem for humankind. Since greenhouses are prone to be the one supply of contemporary meals for Mars area crews and settlers, improvement of synthetic intelligence (AI) and laptop imaginative and prescient primarily based applied sciences for plant progress automation is perceived as a precedence analysis goal. A check web site is already in place for growing and testing superior life assist methods: an autonomous plant cultivation module is working on the Antarctic Neumayer Station III close to the South Pole. Proper now scientists concentrate on creating an AI system that might accumulate details about all of the plant progress components and seedling well being and management greenhouses in autonomous mode with out human involvement.
“One can’t keep steady communication with Neumayer III, whereas coaching laptop imaginative and prescient fashions on board requires too many assets, so we needed to discover a approach to ship a stream of plant pictures to exterior servers for information processing and evaluation,” Skoltech PhD pupil Sergey Nesteruk explains.
As a conclusion to their analysis, the Skoltech staff processed a group of photographs from distant automated methods utilizing their new strategy primarily based on convolutional neural networks and outperforming fashionable codecs by over 7 instances in lowering the picture measurement with out obvious high quality degradation. The researchers used the data from the reconstructed photographs to coach a pc imaginative and prescient algorithm which, as soon as educated, is able to classifying 18 plant varieties in keeping with species at completely different phases of improvement with an accuracy of 92%. This strategy makes it doable to each visually monitor the system operation and constantly collect new ML mannequin coaching information so as to improve the fashions’ performance.
There are plans to deploy and check the brand new methods proper on Neumayer III, which can mark an essential step in the direction of automation of plant rising modules, thus eradicating one more roadblock on the best way to Mars.
Skoltech is a non-public worldwide college positioned in Russia. Established in 2011 in collaboration with the Massachusetts Institute of Expertise (MIT), Skoltech is cultivating a brand new era of leaders within the fields of science, know-how and enterprise, is conducting analysis in breakthrough fields, and is selling technological innovation with the purpose of fixing essential issues that face Russia and the world. Skoltech is specializing in six precedence areas: information science and synthetic intelligence, life sciences, superior supplies and trendy design strategies, vitality effectivity, photonics and quantum applied sciences, and superior analysis.
Disclaimer: AAAS and EurekAlert! are usually not chargeable for the accuracy of stories releases posted to EurekAlert! by contributing establishments or for using any info by way of the EurekAlert system.