Coin Sorting Robot

Project Overview

Built during an intense 24-hour hackathon, this coin sorting robot uses machine learning to identify different Indian coin denominations (₹1, ₹2, ₹5, ₹10) and automatically sorts them into separate containers. The entire system runs on a Raspberry Pi 4B.

The Challenge

Creating a working coin sorting system in just 24 hours required:

• Training a machine learning model to recognize coins
• Building physical sorting mechanism
• Integrating camera system for coin detection
• Designing reliable mechanical components
• Testing and debugging under time pressure

System Architecture

The coin sorting system combines machine learning with mechanical engineering to create an efficient automated solution:

• Camera captures high-quality images of coins
• ML model processes images for denomination identification
• Control system directs coins to appropriate containers
• Mechanical system handles physical sorting operations

Processing Power

The Raspberry Pi 4B provides the computational power needed for real-time coin recognition:

• Runs machine learning inference in real-time
• Processes camera feed at optimal frame rates
• Controls servo motors for precise sorting
• Manages the entire sorting workflow efficiently

Mechanical Design

The sorting mechanism was designed for reliability and precision during the 24-hour hackathon:

• Four separate containers for different denominations
• Servo-controlled gates for precise coin direction
• Smooth conveyor system for continuous operation
• Robust construction using available materials

How It Works

• Camera captures images of coins on the conveyor
• Machine learning model identifies coin denomination
• System calculates sorting path for each coin
• Mechanical actuators direct coins to correct containers
• Process repeats for continuous sorting operation

AI Recognition System

The machine learning component was crucial for accurate coin identification under time constraints:

• Data Collection: Rapid gathering of training images for ₹1, ₹2, ₹5, and ₹10 coins
• Preprocessing: Image enhancement and normalization for better recognition
• Model Training: Convolutional neural network optimized for coin classification
• Performance: Achieved high accuracy with fast inference on Raspberry Pi

24-Hour Hackathon Challenges

• Time Constraint: Prioritized core functionality and built minimal viable product first
• Model Training: Used transfer learning to quickly train accurate coin recognition model
• Hardware Integration: Rapid prototyping with available materials for quick assembly
• Real-time Performance: Optimized code for smooth operation on Raspberry Pi

Results

Successfully demonstrated a working coin sorting system within the 24-hour deadline. The robot accurately identified and sorted all four Indian coin denominations with high precision, showcasing the power of combining machine learning with mechanical engineering.

Key Learnings

• Rapid prototyping and time management under pressure
• Practical machine learning implementation on edge devices
• Integration of AI with physical mechanical systems
• Real-time image processing optimization
• Hardware-software integration for robotics applications