Precision agriculture has changed the way farmers manage their business, with the use of various technologies, such as drones for scouting fields and aerial imagery, and artificial intelligence and robotics. However, development of image analysis tools for classification and precise estimation of plant features is extremely challenging, especially on grassland canopies. While available tools are mostly effective on row- and horticultural crops, analyzing the highly complex vegetation environment of grass swards, including leaves overlap, is a problem yet to be solved. Farmers require recommendations to cope with production and climate challenges, while improving sustainability and protecting the environment. Consequently, a growing demand for “actionable intelligence” is visible among grassland producers. To address this need, a new image analysis tool (i.e., solution), Grasision, is being developed at NIBIO, to estimate key indicators of sward quality status, namely plant and bare soil coverages and leaf area index -none of which is supported by existing solutions. Our tool will provide farmers with critical input to take cost-effective and responsible measures for fertilization, sward renovation and harvesting. Grasision results from devising a new image analysis and classification algorithm, using machine learning and artificial intelligence techniques, to automatically process color images acquired via on-ground platforms or drones. The main objective of this project is to improve the tool’s segmentation and classification accuracy, by refining the algorithm with new training data and ground-truth samples taken on new grass fields. Moreover, we will develop the software platform, and work actively to secure applicable IP rights and industrial partnerships. Additionally, this project sets the foundation for a subsequent funding application for a commercialization project, while supporting NIBIO's ongoing research on precision agriculture and forage production.
Project leader: Izaskun Muruzábal-Lecumberri
Institution: VALIDÉ AS