Computer Vision System for Mango Fruit Defect Detection Using Deep Convolutional Neural Network.

Area of Science:

• Agricultural Technology
• Computer Vision
• Machine Learning

Background:

• Manual fruit grading is subjective and inefficient, impacting export quality.
• Automated systems are crucial for improving agricultural productivity and economic growth.
• Defect detection on fruit surfaces is vital for the export market, especially for popular fruits like mangoes.

Purpose of the Study:

• To develop a computer-assisted grading system for defect detection in mangoes.
• To implement a machine learning approach for automated mango quality classification.
• To enhance the efficiency and consistency of mango grading for the export market.

Main Methods:

• A computer-vision system utilizing convolution neural networks (CNNs) was developed.
• The system was trained and tested using a publicly available mango image database.
• Deep learning techniques were employed for digital image classification and defect detection.

Main Results:

• The proposed CNN-based system achieved a high accuracy of 98% in classifying mango quality.
• Experimental results demonstrated the effectiveness of the automated system.
• The system successfully identified defects on mango surfaces.

Conclusions:

• The developed computer-vision system provides an accurate and efficient solution for mango quality grading.
• Machine learning, specifically CNNs, offers a powerful tool for automating agricultural quality assessment.
• This automated approach can significantly improve the consistency and speed of fruit inspection for export markets.