Artificial Vision Systems for Fruit Inspection and Classification: Systematic Literature Review

Area of Science:

• Agricultural Engineering
• Computer Vision
• Food Science

Background:

• Computer vision is crucial for ensuring quality and safety in the fruit industry.
• Automated fruit sorting systems require efficient and accurate inspection techniques.
• Existing literature covers various aspects but lacks a consolidated overview of current practices.

Purpose of the Study:

• To systematically review and identify applications, hardware, and algorithms for computer vision-based fruit sorting.
• To provide a comprehensive overview of the state-of-the-art in fruit quality inspection.
• To guide the development of advanced fruit classification systems.

Main Methods:

• Systematic literature review following the PRISMA methodology.
• Analysis of 56 articles published between 2015 and 2024 from Web of Science and Scopus.
• Categorization of findings based on application areas, hardware configurations, and processing techniques.

Main Results:

• Key application areas identified: orchards, industrial processing, and retail/home environments.
• Predominant hardware includes RGB cameras and LED lighting; multispectral cameras are vital for complex tasks like foreign material detection.
• Common processing techniques range from traditional algorithms (Otsu, Sobel) to deep learning models (ResNet, VGG), often utilizing transfer learning.

Conclusions:

• The study provides a foundational guide for developing fruit quality inspection and classification systems.
• Diverse environments necessitate tailored hardware and algorithmic approaches for optimal performance.
• Advancements in computer vision, particularly deep learning, are transforming fruit sorting and quality assessment.