Classification of oolong tea varieties based on computer vision and convolutional neural networks

Area of Science:

• Food Science and Technology
• Computer Vision
• Machine Learning

Background:

• Accurate differentiation of oolong tea varieties is crucial for traceability and quality control in the food industry.
• Current methods for oolong tea classification face challenges in speed and accuracy.
• Automated classification using visual attributes is needed to address these limitations.

Purpose of the Study:

• To develop an automated system for classifying oolong tea varieties using machine learning and computer vision.
• To evaluate the performance of Support Vector Machines (SVM) and Convolutional Neural Networks (CNN) for this task.
• To identify key visual characteristics for oolong tea differentiation.

Main Methods:

• Extraction of 13 visual characteristics (color, texture) from five oolong tea varieties.
• Implementation of data augmentation and image cropping to enhance model robustness.
• Development and comparison of SVM and CNN models, including the ResNet50 architecture.

Main Results:

• The ResNet50 model, a type of Convolutional Neural Network (CNN), achieved a Top-1 accuracy exceeding 93% in oolong tea classification.
• Visual attributes like color and texture were successfully used to differentiate between tea varieties.
• The study demonstrated the effectiveness of the implemented machine learning models.

Conclusions:

• Integrating computer vision with machine learning offers a promising, non-invasive approach for rapid and accurate oolong tea categorization.
• This methodology has significant implications for process monitoring, quality assurance, and authenticity validation in the tea industry.
• The findings support the use of automated visual analysis for detecting adulteration and ensuring tea product integrity.