Related Experiment Video
Updated: Aug 1, 2025
04:21
Author Spotlight: Sieving Fruit Pulp to Detect Immature Tephritid Fruit Flies in the Field
Published on: July 28, 2023
1.8K
A Dragon Fruit Picking Detection Method Based on YOLOv7 and PSP-Ellipse
Jialiang Zhou1,2, Yueyue Zhang1,2, Jinpeng Wang1,2
1School of Mechanical and Electronic Engineering, Nanjing Forestry University, Nanjing 210037, China.
Sensors (Basel, Switzerland)
|April 28, 2023
Summary
This study introduces a new method for detecting dragon fruit, including its endpoints, to aid automated harvesting. The approach enhances robotic picking by providing crucial visual data for complex fruit postures.
Area of Science:
- Agricultural Engineering
- Computer Vision
- Robotics
Background:
- Manual dragon fruit harvesting is labor-intensive due to the fruit's complex postures and tough branches.
- Automated picking of dragon fruit is challenging because existing methods struggle with diverse fruit orientations.
Purpose of the Study:
- To develop an advanced dragon fruit detection system for automated harvesting robots.
- To accurately identify and locate dragon fruits, and detect their head and root endpoints for improved robotic manipulation.
Main Methods:
- Utilized YOLOv7 for initial dragon fruit detection and classification.
- Proposed a PSP-Ellipse method combining PSPNet for segmentation, ellipse fitting for endpoint positioning, and ResNet for endpoint classification.
- Validated the method through experiments on dragon fruit detection, segmentation, and endpoint analysis.
Main Results:
- YOLOv7 achieved high precision (0.844), recall (0.924), and average precision (0.932) in dragon fruit detection.
- PSPNet demonstrated superior segmentation performance with precision (0.959), recall (0.943), and mIoU (0.906).
- The PSP-Ellipse method showed improved endpoint detection accuracy compared to regression-based methods.
Conclusions:
- The proposed dragon fruit detection and endpoint identification method effectively supports automated harvesting.
- This research offers a valuable reference for developing fruit detection systems in other agricultural applications.
