Related Experiment Video
Updated: Dec 6, 2025
09:41
Estimation of Contact Regions Between Hands and Objects During Human Multi-Digit Grasping
Published on: April 21, 2023
2.0K
Real-Time Fruit Recognition and Grasping Estimation for Robotic Apple Harvesting
Hanwen Kang1, Hongyu Zhou1, Xing Wang1
1Laboratory of Motion Generation and Analysis, Faculty of Engineering, Monash University, Clayton, VIC 3800, Australia.
Sensors (Basel, Switzerland)
|October 6, 2020
Summary
A new deep learning vision system enables autonomous apple harvesting with high accuracy and efficiency. This robotic system achieves an 0.8 success rate, overcoming traditional vision method limitations in agriculture.
Area of Science:
- Agricultural Robotics
- Computer Vision
- Deep Learning
Background:
- Traditional vision systems struggle with accuracy, robustness, and efficiency in real-world agricultural harvesting.
- Developing autonomous robotic harvesting requires robust and precise vision capabilities.
Purpose of the Study:
- To develop and evaluate a fully deep learning-based vision method for autonomous apple harvesting.
- To address the limitations of traditional vision methods in robotic harvesting applications.
Main Methods:
- A light-weight one-stage detection and segmentation network was used for fruit recognition from RGB-D camera inputs.
- A PointNet model processed point clouds and RGB image data to estimate fruit approach poses for robotic grasping.
- The system was evaluated using RGB-D image data from laboratory and orchard environments.
Main Results:
- The developed vision method demonstrated high efficiency and accuracy in guiding robotic harvesting.
- Robotic harvesting experiments achieved a 0.8 success rate.
- The system achieved a cycle time of 6.5 seconds per harvest.
Conclusions:
- The fully deep learning-based vision method is effective for autonomous apple harvesting.
- The developed system overcomes challenges in accuracy, robustness, and efficiency compared to traditional methods.
- This approach shows significant promise for the future of agricultural automation.
