Detection of Green Asparagus in Complex Environments Based on the Improved YOLOv5 Algorithm

Area of Science:

• Computer Vision
• Agricultural Robotics
• Machine Learning

Background:

• Intelligent machine harvesting of green asparagus requires efficient and accurate detection systems.
• Existing algorithms may struggle with accuracy in complex environments and varying weather conditions.

Purpose of the Study:

• To develop an improved YOLOv5 algorithm for accurate asparagus recognition and detection.
• To enhance the algorithm's performance in complex environments for intelligent machine harvesting.

Main Methods:

• Incorporated a coordinate attention (CA) mechanism into the YOLOv5 backbone.
• Replaced PANet with BiFPN in the neck for improved feature propagation and reuse.
• Constructed a diverse dataset of asparagus under various weather conditions.

Main Results:

• The improved YOLOv5 model achieved a mean Average Precision (mAP) of 98.69% at an IoU threshold of 0.5.
• This represents a 4.22% increase in mAP compared to the original YOLOv5 prototype.
• The enhanced model demonstrated effective detection capabilities in complex and diverse conditions.

Conclusions:

• The proposed improved YOLOv5 algorithm offers high accuracy for asparagus detection.
• It provides robust technical support for intelligent machine harvesting of asparagus across different weather conditions and complex environments.