Granite classification using machine learning and edge computing.

Area of Science:

• Materials Science
• Computer Science

Background:

• Interior decor and ambiance heavily rely on design materials like granite.
• A system is needed to help consumers differentiate granite types and prevent fraudulent substitutions by merchants.
• Existing solutions using Convolutional Neural Networks (CNNs) and image processing lack precision and computational efficiency.

Purpose of the Study:

• To develop a precise and computationally efficient granite classification system.
• To aid end-users in selecting granite that complements their home decor.
• To prevent fraud by ensuring merchants deliver the exact granite color selected by the consumer.

Main Methods:

• Developed a machine learning-based granite classifier utilizing Edge Computing.
• Integrated a color sensor (TCS3200) with an ESP8266 board to acquire RGB data from rocks.
• Trained a machine learning algorithm, specifically Random Forest, using acquired color data for granite classification.

Main Results:

• The system achieved 94% accuracy in classifying granite types.
• The developed system provides a reliable method for differentiating various granite stones.
• Edge computing integration ensures computational efficiency for real-time classification.

Conclusions:

• The proposed system empowers end-users to confidently distinguish between different granite types.
• This technology offers a practical solution for verifying granite authenticity in commercial transactions.
• The system enhances consumer trust and satisfaction in granite selection and purchase.