Bachelor's Thesis Project developed at the Petroleum-Gas University of Ploiești, Faculty of Letters and Sciences, Computer Science Specialization.
TrafficInsights is a complex, distributed software solution designed to monitor and analyze road traffic in real-time. By leveraging State-of-the-Art Deep Learning models and a robust distributed architecture, the system provides high-accuracy traffic metrics essential for smart city infrastructure and road management.
- Real-time Object Detection: Uses YOLOv26 (Medium) to identify and classify vehicles (cars, trucks, buses, motorcycles).
- Advanced Tracking: Implements ByteTrack and Kalman Filters to maintain vehicle identity across frames, even during partial occlusions.
- Precise Speed Estimation: Applies Perspective Transformation to map 2D image coordinates to 3D real-world ground coordinates for accurate speed calculation.
- Live Video Streaming: Low-latency video delivery using FFmpeg and HLS (HTTP Live Streaming).
- Comprehensive Analytics:
- Average & Median Speed, 85th Percentile Speed.
- Traffic Intensity and Density.
- Average Headway (time between consecutive vehicles).
- Historical data visualization via interactive charts.
The project features a distributed architecture across two different computing environments, connected via a Secure SSH Tunnel (Port Forwarding).
| Component | Environment | Tech Stack |
|---|---|---|
| Computer Vision Engine | Linux (NVIDIA GPU) | Python, YOLOv26, ByteTrack, OpenCV, Supervision |
| Streaming Server | Linux | FFmpeg, HLS, Python (SimpleHTTPServer) |
| Backend API | Windows | ASP.NET Core, Entity Framework Core, SQL Server Express |
| Frontend Dashboard | Web Browser | Angular, TypeScript, Bootstrap, Ngx-Charts |
- Artificial Intelligence: YOLOv26 (Object Detection), ByteTrack (Multi-Object Tracking).
- Backend: C# with ASP.NET Core (REST API), Identity for Secure Authentication.
- Frontend: Angular 17+, TypeScript, RxJS for real-time data updates.
- Database: Microsoft SQL Server (Relational storage for traffic events and metrics).
- DevOps/Networking: SSH Tunneling, FFmpeg for video transcoding.
For a quick overview of the entire work, please refer to the project poster: View Thesis Poster
The full technical documentation (Thesis book) containing all theoretical foundations, implementation details, and test results can be found in the /Docs folder.
Vasile Andrei - Daniel
Computer Science,
Petroleum-Gas University of Ploiești,
2026