Smart-Traffic-Signal

Smart traffic signals are designed to optimize the flow of vehicles and pedestrians at intersections. These systems adjust signal timings dynamically based on real-time traffic data, ensuring efficient and safe movement through intersections.

This project focuses on implementing a simplified smart traffic signal system. It integrates features like car detection, adaptive green light timing based on traffic flow, and a pedestrian signal.

For simplicity, the model simulates a single intersection with consistent traffic flow and car speeds. The images and videos below showcase the final model in action.

---

Key Features

1. Dynamic Green Light Timing
   Adaptive signal timings based on the number of cars detected in each lane.

2. Car Detection Using OpenCV
   The system detects cars (represented by red paper rectangles) using color-based image processing.

3. Pedestrian Signal Integration
   Pedestrian signals operate with a buzzer and a button for user activation.

4. Linear Regression Optimization
   Dummy traffic data is analyzed to determine optimal green light duration, minimizing congestion.

---

Project Breakdown

Step 1: State Machine Design

A state machine was created to handle the following states:

• Red Light
• Green Light
• Yellow Light
• Pedestrian Signal

Transitions occur in the sequence Red → Green → Yellow → Red, with interruptions for pedestrian signals as needed.

Timings:

• Red and Yellow: Fixed (5 seconds each)
• Green: Dynamic, based on car count
• Pedestrian Signal: Fixed (20 seconds)

---

Step 2: Car Detection

Using OpenCV, the system detects red rectangles (representing cars) on a cardboard intersection model. The process includes:

• Converting the image from BGR to HSV color space
• Applying masks to detect red objects
• Counting the detected "cars" to adjust green light timing

---

Step 3: Data Analysis and Timing Optimization

A dummy dataset was created to simulate traffic flow. Using linear regression, the optimal green light duration (y = 1.47x + 8.43) was calculated based on car count (x) and the time (t) it took for them to get across. In the real world, engineers would have to make these datasets using observations for the intersection they are analyzing.

---

Step 4: Model Intersection Creation

A 12x12-inch cardboard model was built to simulate the intersection, featuring lanes, signals, and cars. The camera was mounted 7 inches above the center for optimal detection coverage as it sees only the lanes and nothing else to reduce any distractions.

---

Step 5: Raspberry Pi Integration

The Raspberry Pi handles signal control, including:

• LEDs for traffic lights
• A camera for detection of cars and pedestrians
• A button and buzzer for pedestrian signal activation

Key commands for setup:

sudo apt update
sudo apt install python3-libraryName

---

Challenges

1. Car Detection Accuracy

   • The initial detection algorithm struggled to differentiate cars from the background.
   • Adding a smaller, controlled intersection model with a consistent background (brown cardboard) improved detection.

2. Time Constraints

   • Limited project time meant focusing on core features, such as car detection and state machine logic, while deferring more advanced functionality like detecting multiple car colors. This also meant that I couldn't finish troubleshooting with the wiring to the raspberry pi.

3. GPIO Power Issue

   • During initial testing, GPIO pins provided power without the expected code execution. Debugging this required reconfiguration and additional testing.

4. Dataset Limitations

   • The YOLO model was unsuitable due to the lack of sufficient labeled data and training time.
   • Opted for OpenCV to detect red cars instead of training a custom neural network which caused there to be less model pieces like pedestrians as paper clips to be used for the project.

Future Improvements

1. Detection Algorithm

   • Detect pedestrians as paper clips and other colored rectangles as different cars.

2. Linear Regression Model for Pedestrians

   • Based on pedestrian detection, I could have made the time for the pedestrian signal longer if my model detects someone with a disability that may need more time to walk. I could have done the same for different types of cars by labeling them with a certain color of paper to add more complexity to the project.

3. Pedestrian Signal Integration

   • Added a pedestrian signal with a button and buzzer for user interaction, despite time constraints.

---