Vision-Based Ripeness Grading of Banana Using YOLO Object Classification

Chok, Yu Hang (2026) Vision-Based Ripeness Grading of Banana Using YOLO Object Classification. Final Year Project (Bachelor), Tunku Abdul Rahman University of Management and Technology.

Text CHOK YU HANG_Full Text.pdf Restricted to Registered users only Download (5MB)

Abstract

Banana ripeness classification is important in post-harvest handling, retail quality monitoring, and supermarket display because ripeness directly influences product quality, market value, shelf management, and consumer acceptance. However, conventional manual inspection is often subjective, inconsistent, and easily affected by variations in lighting and background conditions, which reduce the reliability of ripeness assessment in practical environments. To address this problem, this project developed a vision-based banana ripeness classification system using the YOLOv8 classification model and compared three image input methods, namely Raw, Masked + CLAHE, and Isolated + CLAHE, under a fair and consistent training setting. The developed system was evaluated through both offline dataset testing and real-time webcam-based testing under warm and cool lighting conditions using a custom GUI platform. The results showed that all three methods achieved similarly high performance in offline evaluation, with Method 2 producing the highest offline accuracy. However, real-time testing revealed clearer practical differences, where Method 3 consistently achieved the highest accuracy, strongest stability, and best robustness under varying lighting conditions. Overall, the study demonstrates that YOLOv8 with suitable image preprocessing provides an effective and practical solution for automated banana ripeness classification, while supporting more sustainable food quality monitoring aligned with SDG 9 and SDG 12