Automated Real-Time Sugarcane Node Detection System Using YOLOv8 and Edge Computing

Abstract

Sugarcane farming is a vital agricultural sector that faces major challenges due to its heavy reliance on manual labor, especially during node identification and harvesting. Traditional methods are labor-intensive, time-consuming, and prone to human errors, leading to decreased productivity and increased operational costs. This research presents an automated, real-time sugarcane node detection and cutting system leveraging deep learning and computer vision technologies. The proposed system utilizes the YOLOv8 object detection model for accurate node identification and integrates actuator control mechanisms to automate cutting. A dataset comprising over 4000 labeled sugarcane stem images was used to train the detection model, achieving an average precision of ~92% and an F1 score of ~90.5%. The solution is deployed on a low-power edge device (Jetson Nano) using Python, OpenCV, and TensorFlow/Keras, enabling efficient real-time performance with an average processing speed of 20–25 FPS. Servo motors are employed for executing precise cuts based on model predictions. This integration significantly reduces dependency on manual labor, enhances cutting precision, and promotes sustainable farming practices. The system\'s modularity, low cost, and adaptability suggest a strong potential for commercial adoption and future expansion into broader smart farming applications.

Introduction

Sugarcane is an important cash crop primarily grown in tropical regions, with manual farming methods still widely used, especially for identifying and cutting stem nodes critical for planting and regrowth. Manual node detection is labor-intensive, error-prone, and limited by labor shortages, driving the need for automation.

This research proposes a real-time automated sugarcane node detection and cutting system using the YOLOv8 deep learning model deployed on a low-power NVIDIA Jetson Nano edge device. The system captures live video, detects sugarcane nodes with high accuracy, predicts the optimal cutting moment, and controls a servo motor to perform precise cuts. This reduces manual labor, improves efficiency, and supports sustainable smart farming.

The paper reviews related work showing prior efforts in agricultural automation but notes a gap in real-time sugarcane node detection and cutting automation. Unlike earlier methods relying on traditional image processing or non-real-time systems, this approach integrates deep learning, edge computing, and mechanical actuation into a practical, deployable solution.

Methodology includes dataset creation with 4000+ annotated images, YOLOv8 model training, real-time detection and cutting prediction, servo control for cutting, and a user interface for system monitoring. Testing validated the system’s accuracy and responsiveness in real conditions.

Future improvements planned include multi-angle cameras, self-learning models, enhanced cutting precision, AI-based quality assessment, remote monitoring dashboards, and sustainable power options.

Conclusion

The developed system successfully addresses the longstanding challenges of manual sugarcane node detection and cutting by introducing real-time automation through machine learning and computer vision technologies. By leveraging YOLOv8 for node identification and integrating actuator control for precise cutting, the project demonstrates significant improvements in efficiency, accuracy, and resource optimization compared to traditional methods. The deployment on a low-power edge device like Jetson Nano ensures portability and cost-effectiveness, making it accessible to small and mid-scale farmers. Extensive testing validated the system’s reliability, achieving high detection accuracy and operational stability under real-world conditions. Furthermore, the modular design allows easy updates and scalability, laying the foundation for future enhancements such as multi-crop adaptability, IoT integration, and mobile field deployment. This project represents a critical step toward the modernization of agriculture, promoting smarter and more sustainable farming practices to meet growing global demands.

References

[1] R. Ehsani, et al., \"Image Processing System for Sugarcane Row Detection,\" Journal of Agricultural Engineering Research, 2013. [2] J. Zhang, et al., \"Support Vector Machine-Based Sugarcane Disease Detection,\" International Journal of Agricultural Sciences, 2017. [3] A. Kumar, et al., \"Growth Stage Identification in Sugarcane Using CNN and UAV Imagery,\" IEEE Transactions on Geoscience and Remote Sensing, 2020. [4] Ultralytics, \"YOLOv8 Official Documentation,\" Available: https://docs.ultralytics.com/ [5] OpenCV Development Team, \"OpenCV: Open Source Computer Vision Library,\" Available: https://opencv.org/ [6] TensorFlow Developers, \"TensorFlow: Open Source Machine Learning Framework,\" Available: https://www.tensorflow.org/ [7] NVIDIA Corporation, \"Jetson Nano Developer Kit User Guide,\" Available: https://developer.nvidia.com/embedded/jetson-nano-developer-kit [8] Roboflow, \"Image Labeling and Annotation for Machine Learning,\" Available: https://roboflow.com/ [9] CVAT, \"Computer Vision Annotation Tool (CVAT) User Manual,\" Available: https://opencv.github.io/cvat/ [10] Raspberry Pi Foundation, \"RPi.GPIO Python Library Documentation,\" Available: https://sourceforge.net/p/raspberry-gpio-python/wiki/Home/

Copyright

Copyright © 2025 Amruta Patil, Samiksha Udagave, Shradha Patil, Aparna Thorvat, Netrali Patil, Prof. U. A Nuli. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.