Back to Projects
Computer Vision Based Approach for Real-Time Road Traffic Violation Detection

Computer Vision Based Approach for Real-Time Road Traffic Violation Detection

University
University of Sri Jayewardenepura•2016 - 2017•Undergraduate Capstone Project

Technologies

pythonc++opencvtensorflow

Key Highlights

  • ▸Developed a real-time computer vision system for automated traffic violation detection using CCTV footage
  • ▸Published first-author research at the 13th International Research Conference, General Sir John Kotelawala Defence University (2020)
  • ▸Contributed as second author to earlier work on automated traffic monitoring for complex road conditions (2018)
  • ▸Designed machine learning–assisted algorithms for detecting speed, signal, and lane violations
  • ▸Implemented a full end-to-end pipeline using OpenCV and TensorFlow

Overview

Undergraduate capstone project focused on real-time traffic violation detection using computer vision and machine learning on CCTV footage.
The system was developed to reduce reliance on manual traffic monitoring and demonstrate the feasibility of automated, scalable traffic enforcement systems, in collaboration with LIRNEasia.

Problem Context

Manual traffic enforcement is:

  • Labor-intensive and costly
  • Error-prone and difficult to scale
  • Limited in coverage and consistency

The project aimed to design an automated system capable of:

  • Processing CCTV feeds in real time
  • Detecting multiple classes of traffic violations
  • Handling complex road layouts and varying environmental conditions

Technical Approach (Summary)

Computer Vision Pipeline

  • Vehicle detection and multi-object tracking
  • Lane and road boundary detection
  • Traffic signal state recognition
  • Temporal analysis across video frames to reduce false positives

Violation Detection

  • Speed violation detection via tracking and distance estimation
  • Red-light violation detection using signal–vehicle correlation
  • Lane discipline and wrong-lane detection
  • Static object analysis for illegal parking
  • Directional analysis for wrong-way driving

Machine Learning Integration

  • Custom annotated traffic datasets
  • Vehicle and object classification models
  • Traffic sign recognition using deep learning
  • Model fine-tuning for local road conditions

Results & Impact

  • Demonstrated real-time violation detection on CCTV footage
  • Achieved robust performance across varying lighting and road conditions
  • Validated feasibility of automated traffic monitoring for urban deployment
  • Research outputs published in peer-reviewed conference proceedings
  • Open-source implementation enabled reproducibility and further research
Computer Vision Based Approach for Real-Time Road Traffic Violation Detection 1

Resources & Artifacts

Source Code

[1] Traffic Violation Detection System (GitHub)
Akila Peiris. Computer vision–based real-time traffic violation detection system.
https://github.com/akilapeirisz/traffic-violation-detection-system

Publications

[2] A. Peiris, E. Edirisuriya, C. Athuraliya, I. Jayasooriya.
Computer Vision Based Approach for Traffic Violation Detection.
Proceedings of the 13th International Research Conference,
General Sir John Kotelawala Defence University, 2020.

[3] R. Opatha, A. Peiris, D. Gamini, E. Edirisuriya, C. Athuraliya, I. Jayasooriya.
Automated Traffic Monitoring for Complex Road Conditions.
2018.