Estimating and evaluating the rice cluster distribution uniformity with UAV-based images

Area of Science:

• Agricultural Engineering
• Remote Sensing
• Computer Vision

Background:

• Uniformity of rice plant distribution is crucial for crop yield and efficient management of agricultural inputs like fertilizers and pesticides.
• Current methods for assessing rice plant distribution uniformity are lacking in technical sophistication and accuracy.

Purpose of the Study:

• To develop and validate a novel technical system for estimating and evaluating rice plant distribution uniformity using unmanned aerial vehicle (UAV) imagery.
• To provide a precise, convenient, and cost-effective solution for agricultural uniformity assessment.

Main Methods:

• Rice cluster recognition and localization using RGB color characteristics from aerial images.
• Region segmentation based on Voronoi diagrams to delineate rice clusters.
• Development of a uniformity index using the coefficient of variation for distribution evaluation.

Main Results:

• High precision in rice cluster recognition (>95% precision, 97% accuracy, 97% recall, 96% F1-score).
• Minimal rice cluster location error (mean error 1.62 cm), following a gamma distribution.
• The proposed uniformity index effectively evaluates distribution, with the overall method showing high accuracy (relative error <0.01% vs. manual labeling).

Conclusions:

• The UAV-based method is a feasible, accurate, and cost-effective tool for assessing rice plant distribution uniformity.
• The technology demonstrates significant potential for precision agriculture and crop management.
• Evaluation highlights areas for improvement in mechanized transplanting uniformity in South China.