Related Experiment Video
Updated: May 26, 2026
11:49
Cereal Crop Ear Counting in Field Conditions Using Zenithal RGB Images
Published on: February 2, 2019
Enhancing precision harvesting in smart orchards: a light-weight neural network for apple maturity detection
Linna Hu1, Penghao Xue1, Weixian Zha1
1School of Network and Communication Engineering, Jinling Institute of Technology, Nanjing, China.
Frontiers in Plant Science
|May 25, 2026
Summary
We developed HRLN-YOLO, a lightweight apple maturity detection model for smart agriculture. It achieves high accuracy with reduced network weights, enabling efficient deployment on edge devices for automated harvesting.
Area of Science:
- Computer Vision
- Machine Learning
- Agricultural Technology
Background:
- Deep learning models for apple maturity detection are crucial for smart agriculture.
- Deployment on edge devices is challenging due to resource constraints, weather, and fruit clustering.
Purpose of the Study:
- To propose HRLN-YOLO, a lightweight and accurate apple maturity detection model for edge deployment.
- To minimize network weights and computational complexity while maintaining high detection performance.
Main Methods:
- Designed a lightweight backbone (HGBackbone) for enhanced feature extraction and faster inference.
- Developed an enhanced neck module (RCF_Neck) for improved multi-scale feature fusion, especially under occlusion.
- Introduced a lightweight detection head (LADH-Head) to reduce task conflicts and computational cost.
- Implemented NWD-Loss to enhance localization stability for small targets.
Main Results:
- HRLN-YOLO achieved a 1.7% improvement in mAP@0.5 compared to the YOLO11n baseline.
- The model demonstrated a 37.3% reduction in parameters and a 34.9% decrease in computational complexity.
- Experiments were conducted on the Orchard Apple Maturity Dataset.
Conclusions:
- HRLN-YOLO offers a practical solution for edge deployment in smart orchards, balancing network efficiency and detection accuracy.
- The model facilitates automated harvesting by providing precise apple maturity detection on resource-constrained devices.
