Leaf-Movement-Based Growth Prediction Model Using Optical Flow Analysis and Machine Learning in Plant Factory.

Area of Science:

• Agricultural Science
• Computer Vision
• Plant Physiology

Background:

• Productivity stabilization in plant factories is crucial.
• Current growth prediction methods using projected plant area (PA) have limitations.
• Accurate growth prediction is key to improving crop yields in controlled environments.

Purpose of the Study:

• To develop a novel method for predicting lettuce fresh weight at harvest.
• To utilize time-series leaf movement analysis via optical flow (OF) for growth prediction.
• To evaluate the performance of OF-based features in a machine learning model for plant growth forecasting.

Main Methods:

• Applied optical flow (OF) analysis to extract leaf movement features from time-series images of lettuce seedlings.
• Collected images of 338 lettuce seedlings every 20 minutes over 9 days using a microcomputer and camera module.
• Utilized machine learning with OF-derived features to predict fresh weight at 38 days post-sowing.

Main Results:

• Achieved a high correlation ratio of 0.743 for the growth prediction model based on OF analysis.
• Demonstrated the effectiveness of analyzing time-series leaf movement for accurate plant growth prediction.
• Identified OF analysis as a promising technique for automated plant phenotyping.

Conclusions:

• Optical flow analysis provides a robust method for predicting lettuce fresh weight in plant factories.
• This approach can significantly improve the accuracy of growth prediction models.
• The developed phenotyping system enables widespread application in commercial plant factories for enhanced productivity.