Related Experiment Video
Updated: Jul 30, 2025
07:58
Author Spotlight: Non-Invasive High-Resolution Measurement of Chlorophyll Synthesis During De-Etiolation
Published on: January 12, 2024
854
Detection of Chrysanthemums Inflorescence Based on Improved CR-YOLOv5s Algorithm
Wentao Zhao1,2,3, Dasheng Wu1,2,3, Xinyu Zheng1,2,3
1College of Mathematics and Computer Science, Zhejiang A&F University, Hangzhou 311300, China.
Sensors (Basel, Switzerland)
|May 13, 2023
Summary
This study introduces an improved CR-YOLOv5s model for accurate chrysanthemum flower recognition, enhancing feature representation with an attention mechanism. The enhanced model achieves higher accuracy, aiding in automated flower processing and yield estimation.
Area of Science:
- Computer Vision
- Agricultural Technology
- Machine Learning
Background:
- Accurate flower stage recognition is crucial for yield estimation.
- Complex backgrounds and subtle visual differences challenge current flower detection methods.
Purpose of the Study:
- To improve the accuracy and robustness of flower bud and bloom recognition in chrysanthemums.
- To enhance feature representation in deep learning models for complex agricultural imagery.
Main Methods:
- Proposed an improved CR-YOLOv5s model incorporating a coordinate attention mechanism.
- Integrated RepVGG block structures into the YOLOv5s backbone for enhanced feature extraction.
- Evaluated the model's performance on chrysanthemum images with complex backgrounds.
Main Results:
- The improved CR-YOLOv5s achieved an average accuracy of 93.9%.
- Demonstrated a 4.5% improvement in accuracy compared to the standard YOLOv5s model.
- Showcased enhanced detection accuracy and robustness in challenging conditions.
Conclusions:
- The CR-YOLOv5s model effectively addresses challenges in flower recognition due to complex backgrounds and subtle visual variations.
- This research provides a foundation for automated flower picking, grading, and yield prediction systems.
- The integration of attention mechanisms and advanced network structures significantly boosts detection performance.
