Development and Validation of a YOLO-Based Drone Vision System for Automated Ginger Plant Height Measurement

Abstract

Accurate and timely monitoring of crop growth is essential for improving productivity. Conventional plant height measurement methods, such as manual observation, are labor-intensive and time-consuming. Agricultural automation using computer vision and unmanned aerial vehicles (UAVs) offers a promising alternative for rapid, non-contact, and scalable crop monitoring. This study presents a YOLO-based drone vision system for automated measurement of ginger plant height. The proposed framework integrates YOLO object detection with ArUco marker-based scale calibration to estimate plant heights. A custom ground-truth dataset comprising 204 annotated images was created for plant detection and height measurement. Image augmentation techniques were employed to improve model robustness and generalizability. A comparative evaluation of three YOLO architectures, namely YOLOv5, YOLOv8, and YOLOv11, was conducted to identify the most suitable model for ginger plant height estimation. The proposed system was evaluated using three imaging platforms: an iPhone 14 Pro Max camera, a stationary Siyi A8 Mini drone camera, and a Siyi A8 Mini camera mounted on a drone in flight. A 17 cm ArUco marker was used as the reference scale for height measurement. Experimental results show that the YOLOv8 model achieved the highest performance, attaining a mAP@50 of 98.1%. The system demonstrated consistent performance across all imaging platforms, with an average percentage error of 5.92% in plant height measurement using a flying drone, relative to manual ground-truth measurements. The proposed system combines drone-based imaging and deep learning for automated, low-cost crop monitoring and precision farming. Results demonstrate the framework's effectiveness for scalable plant growth assessment in agricultural applications.