Instant Prediction of the rice yield by Artificial Intelligence-based image analysis
Rice (Oryza spp.) is one of the most important crops supplying up to 20% of the world’s food energy. To face climate change and increase food demand, it is crucial to enhance rice production. However, it is a time and labor-consuming task as it requires harvesting, transportation, drying, separating grains from other parts, and weighing. The process makes it difficult and quasi-impossible to evaluate rice productivity of large number of rice farmers on the experimental station in real time. It is also challenging to evaluate the productivity on farmers’ fields at regional scale.
Dr. Yu Tanaka (Associate Prof., Okayama University, Japan; former Assistant Prof., Kyoto University, Japan), Dr. Kazuki Saito (International Rice Research Institute; former agronomist, Africa Rice Center) and their collaborators developed the Artificial Intelligence (AI)-based image analysis model enabling the instant estimation of rice yield using ground-captured images. The research was published in a preview version on Plant Phenomics journal, June 29th, 2023, (https://doi.org/10.34133/plantphenomics.0073) while the final version will be published on July 28th, 2023.
Firstly, the research team collected a set of rice canopy images and corresponding grain yield across 7 countries. The database consisted of over 20,000 images and 400 farmers with various levels of grain yield, growth environments and plant shape. The AI model was then trained by using this database. The results showed that the developed AI model accurately predicts rice grain yield by using only images of the rice canopy at harvest.
The remarkable feature of this technology is that there is no need to use any special instruments or high-level techniques. The only needed resource is a digital image of the rice canopy, which can be easily obtained by using a digital camera or a smart phone. The AI model developed in this study is deployed to the smartphone application ‘HOJO’ in iOS and Android. This usability will enable the AI model to be adapted to a wide range of fields of crop science and agronomy. Since the technology greatly speeds up yield evaluation, it can accelerate the screening rice breeding lines or tolerance to the stresses in the breeding programs. The productivity of the farmers’ field can be easily evaluated with non-invasive manner, which leads to the impact assessment of productivity-enhancing interventions and detection of fields where these are needed to sustainably increase crop production, especially in global south.
This study was financially supported by the European Union and International Fund for Agricultural Development (IFAD) under the project “Sustainable and Diversified Rice-based Farming Systems [DCI FOOD/2015/360–968]” under the program “Putting Research into Use for Nutrition, Sustainable Agriculture and Resilience (PRUNSAR)”
Press Release issued by Okayama University and Africa Rice Center