Maskininlärning för automatisk stressdetektion i hortikulturell produktion
Purpose and goal: Pests and pathogens is a big problem in intensive greenhouse operation and the purpose of this project is to study if it is possible to develop an algorithm for early biotic stress detection, earlier than what a human eye is capable of, based on data from optical sensors using machine learning as a tool.
Expected results and effects: The method being developed is hopefully applicable in commercial greenhouse operation, so that it can be integrated in Heliospectra’s intelligent lighting system. The long-term benefits for the growers are manifold, including improved energy and resource use efficiency, due to reduced waste, as well as reduced use of pesticides. Approach and implementation: Measurement data from plants with and without an ongoing fungi infection have been collected in a neighboring project with different types of optical sensors. The sensor list includes sensor and camera measuring canopy chlorophyll fluorescence, sensor and camera measuring canopy temperature as well as ordinary RGB-camera. Parts of this data set will be used in the project. In step 1, the method is evaluated and selected and data is treated and processed. In step 2, developed algorithms are trained, validated and tested. Heliospectra is collaborating with Chalmers in this project
Torsten Wik (contact)
Professor at Chalmers, Electrical Engineering, Systems and control, Automatic Control
Funding Chalmers participation during 2020–2021