A longan yield estimation approach based on UAV images and deep learning.

Area of Science:

• Agricultural Science
• Computer Vision
• Machine Learning

Background:

• Accurate longan yield estimation is crucial for market pricing and harvest efficiency.
• Current manual methods are labor-intensive and costly.
• Developing automated yield estimation can significantly benefit longan orchard management.

Purpose of the Study:

• To propose an automated method for estimating longan yield in complex natural environments.
• To leverage deep learning and regression analysis for fruit counting and yield prediction.
• To reduce the labor costs associated with traditional longan yield estimation.

Main Methods:

• Utilized Unmanned Aerial Vehicle (UAV) for collecting mature longan canopy video images.
• Developed the CF-YD and SF-YD models for automated identification of Cluster_Fruits and Single_Fruits.
• Employed regression analysis on detected fruit counts and real orchard data to build estimation models.

Main Results:

• Achieved an average error rate of 2.66% for Cluster_Fruits estimation.
• Achieved an average error rate of 2.99% for Single_Fruits estimation.
• Demonstrated the effectiveness of the proposed deep learning models in complex environments.

Conclusions:

• The developed method provides an effective and efficient solution for longan yield estimation.
• The automated approach can significantly improve harvest efficiency and economic benefits for longan orchards.
• This technology offers valuable guidance for optimizing longan fruit harvesting practices.