National Agriculture and Food Research Organization(NARO) in collaboration with Bandai Namco Research Inc. has developed a method using drones and artificial intelligence(AI) to easily estimate the coverage of legumes in pastures where grasses and legumes are mixedly sown. By this research result, it is possible to evaluate the ratio of legumes with high accuracy and efficiency in about 1/5,000 times of manual image analysis. Therefore, precise grassland management such as fertilization according to the density of legumes and additional sowing of insufficient legumes is possible. It is also expected to be used for the development of varieties of legumes suitable for mixed sowing.
Overview
In pastures where grazing and weeding are carried out, mixed sowing cultivation of grasses and legumes is widely practiced for the purpose of improving feed productivity and quality. In order to maximize its benefits, it is necessary to maintain a proper proportion (about 30%) of legumes in the feed.
NARO and Bandai Namco Research Inc. has jointly developed a vegetation evaluation method that estimates the coverage of legumes on aerial images using an AI model by taking pictures of grasslands mixed with grasses and legumes using a drone. It takes more than 3 hours to manually identify legumes in a 1m2 pasture from aerial images and estimate coverage. However, by using the AI model, it is possible to estimate the coverage automatically at high speed (about 2.5 seconds) with the same accuracy.
It is expected that this research result will enable us to easily grasp the ratio of legumes in the grassland and precise grassland management such as local fertilization and additional sowing is possible. In addition, if the legumes grow vigorously then they suppress the growth of grasses, but on the contrary if they are weak then they will be suppressed. The composition ratio in mixed grassland is an important evaluation item in the breeding of legume varieties. Therefore, this result is expected to contribute to the high accuracy and efficiency of cultivar breeding.
We will realize high yield and high-quality production of grass by grassland management using drone and AI and contribute to the achievement of the Sustainable Development Goals (SDGs) goal "Zero hunger" and simultaneously aim to realize a sustainable society.
Publication
Fujiwara R, Nashida H, Fukushima M, Suzuki N, Sato H, Sanada Y and Akiyama Y (2022) Convolutional Neural Network Models Help Effectively Estimate Legume Coverage in Grass-Legume Mixed Swards. Front. Plant Sci. 12:763479.
For Inquiries
Contact: https://www.naro.go.jp/english/inquiry/index.html
