Smart-Traffic-Analysis-System

🚦 Smart Traffic Analysis System

An end-to-end Computer Vision pipeline for real-time traffic monitoring that combines detection, tracking, and analytics into a single system.

📌 Features 🚗 Vehicle Detection using YOLOv8 🔁 Multi-Object Tracking with ByteTrack 📏 Real-World Speed Estimation using homography mapping 🔄 Wrong-Way Detection based on trajectory direction 🚨 Red-Light Violation Detection using stop-line crossing 🎥 Annotated Video Output with IDs, speed, and alerts ⚡ Performance Monitoring (FPS) 🧠 System Overview

This system processes a video input and performs:

Object detection (vehicles) Object tracking with persistent IDs Pixel → real-world coordinate transformation Speed estimation in km/h Behavior analysis (direction + violations) 🛠️ Tech Stack Python OpenCV Ultralytics YOLOv8 ByteTrack NumPy 📂 Project Structure traffic-analysis/ │── main.py │── input.mp4 │── yolov8n.pt │── bytetrack.yaml │── README.md ⚙️ Installation pip install ultralytics opencv-python numpy ▶️ Usage Place your video file: input.mp4 Run the script: python main.py Press Q to exit. 📐 Homography Calibration (Important)

For accurate speed estimation, you must calibrate:

src_pts = np.float32([...]) # pixel points dst_pts = np.float32([...]) # real-world meters

👉 Tips:

Measure lane width or road distance in meters Match 4 points from image → real-world rectangle 🚨 Violation Detection Logic 🔴 Red Light Violation Vehicle crosses stop line while signal = RED 🔄 Wrong Way Detection Vehicle direction ≠ expected traffic flow ⚠️ Limitations Speed depends heavily on homography accuracy Traffic light is simulated (not detected yet) Works best with stable, fixed camera 🚀 Future Improvements Real traffic light detection (YOLO model) Lane detection using deep learning Live dashboard (Flask / Streamlit) Database logging (violations + stats) Multi-camera system

https://github.com/user-attachments/assets/72a24807-cb94-40e7-98f8-50802a111e78 👤 Author

Awais Shah AI / ML | Computer Vision

⭐ If you like this project

Give it a ⭐ on GitHub and feel free to contribute!