AI monitors daily rice growth, cutting field survey labor by 60%

A system has been established in which cameras and AI replace the manual field surveys that once required workers to wade into paddies to assess rice growth. Under the old method, at least two surveyors had to visit each site from transplanting through harvest, but the new AI-based approach allows daily remote monitoring of growth changes. The Rural Development Administration says the system can cut the labor required for field surveys by more than 60 percent.

The Rural Development Administration announced Tuesday that it has built an "AI rice growth monitoring system" that uses video footage and AI to analyze rice growth and yield, and is verifying its field applicability and effectiveness.

Under the current survey method, workers enter paddies in person to check growth conditions and yield-related traits. Surveys are typically conducted twice a month from transplanting to harvest — nine rounds in total — with at least two people deployed each time. As extreme weather events such as heat waves and heavy rains have grown more frequent, and agricultural disasters including pest damage have become more common, more frequent monitoring has become necessary. Manual field surveys alone, however, have faced limits in terms of staffing and time.

The system, developed by the National Institute of Crop Science under the Rural Development Administration, collects footage from fixed points every day and analyzes it with AI. Surveyors can track daily rice growth without visiting the field. The administration says the remote, unmanned and non-destructive monitoring approach can reduce field survey labor by more than 60 percent compared with the existing method.

The system consists of video collection equipment, an AI model for image-based growth assessment, and a rice growth monitoring platform. The National Institute of Crop Science began building the system in 2023, installing fixed video collection devices at 19 provincial agricultural research and extension stations across the country.

Last year, the institute also developed portable video collection equipment that is easy to install and relocate. This expanded the number of video collection points to 86, including sites at city and county agricultural technology centers.

The range of traits the AI can analyze from video has also grown more precise. The current accuracy rate for planting density — the number of rice hills per unit area — stands at 97 percent, while tiller count, the number of stems per plant, reaches 98 percent. The heading date, when rice ears emerge, is analyzed at 99 percent accuracy. The model that estimates rice yield based on ear area achieves 88 percent accuracy.

The Rural Development Administration plans to raise the accuracy of growth trait analysis to 99 percent and improve the yield estimation model to around 95 percent accuracy. Collected footage and AI analysis results will be managed through the rice growth monitoring platform.

AI's role extends beyond tracking how much the rice has grown. The National Institute of Crop Science has also developed a program that uses drone footage to analyze the extent of lodging — when rice plants fall over — and damage caused by brown planthoppers. The goal is to enable rapid assessment of agricultural disaster damage across wide areas without requiring workers to check each paddy individually.

The system's coverage will also expand significantly. The Rural Development Administration plans to gradually increase the number of video collection sites and apply the system to all 690 national rice growth survey points by 2030. Over the longer term, the administration envisions extending the video and AI technologies developed for rice to other crops, ultimately shifting the entire crop growth survey framework to a video- and AI-centered approach.

"The AI rice growth monitoring system is a core achievement of the agricultural AI transformation — converting the existing labor-intensive survey method into a data-driven one," said Kim Ki-young, head of the Rural Development Administration's Basic Food Crops Division. "We will work to broaden the application of video and AI technology beyond rice growth and yield surveys to areas such as disaster diagnosis, and to ensure that our research results can be put to practical use on actual farms."