Fruit Sizing in Orchard: A Review from Caliper to Machine Vision with Deep Learning

Area of Science:

• Horticulture
• Agricultural Engineering
• Computer Vision

Background:

• Automated fruit and vegetable sizing has transitioned from mechanical methods to machine vision in packhouses over 30 years.
• Current advancements focus on applying these technologies for in-orchard fruit size assessment on trees.

Purpose of the Study:

• To review the state-of-the-art in in-orchard fruit sizing technologies.
• To identify key challenges and future directions for machine vision in fruit size assessment.

Main Methods:

• Review of allometric relationships between fruit weight and lineal dimensions.
• Analysis of traditional and machine vision methods for measuring fruit lineal dimensions.
• Examination of depth measurement and occlusion handling in machine vision systems.
• Discussion of sampling strategies and predictive modeling for fruit size at harvest.

Main Results:

• Summarizes commercially available in-orchard fruit sizing capabilities.
• Highlights challenges in depth measurement and recognizing occluded fruit using machine vision.
• Identifies key areas for future development in automated in-orchard fruit sizing.

Conclusions:

• Machine vision offers significant potential for advancing in-orchard fruit size assessment.
• Further research and development are needed to overcome current limitations and enhance predictive accuracy.