This Project relates to an advanced AI-driven system designed to automate the detection of traffic violations and the issuance of electronic tickets (e-tickets) using computer vision. The system leverages custom Convolutional Neural Network (CNN) architectures, trained on a locally collected and annotated dataset from the Punjab Safe Cities Authority (PSCA), to identify 19 different types of traffic violations. The innovation is divided into two modules: the first module detects violations that can be identified from single image frames, such as helmet detection, seatbelt usage, mobile phone usage, and others. The second module employs temporal analysis to detect violations that require understanding vehicle movement and behavior over time, including over speeding, red signal violations, and wrong way driving. Both modules are optimized for real-time processing and deployed on in-house GPU servers, ensuring efficient and immediate enforcement.

The system’s integration with existing city surveillance infrastructure allows for seamless deployment, providing comprehensive traffic monitoring and enhancing road safety. By utilizing deep learning and computer vision, the system offer high accuracy and reliability tailored to the specific traffic conditions of Pakistan. This automated approach reduces the need for manual monitoring, allowing law enforcement to focus on other critical tasks. The scalable design ensures that the system can be adapted to various surveillance setups and deployed across different cities. This invention represents a significant advancement in traffic law enforcement technology, improving compliance and safety on the roads of Pakistan.

Method

The system comprises two main modules, each designed to address different categories of traffic violations:

Single-Frame Violation Detection Module

This module uses custom CNN architectures to detect violations that can be identified from a single image frame. It includes detection of violations such as:

  1. Helmet detection
  2. Seatbelt usage
  3. Mobile phone usage while driving
  4. Tinted glasses
  5. Overloading in public service vehicles
  6. Goods vehicle overloading
  7. Vehicles without proper lights at night
  8. Zebra crossing, Lane/line violations
  9. Emitting smoke
  10. Unregistered vehicles
  11. Underage drivers
  12. Pillion riding

Temporal Violation Detection Module

This module focuses on violations requiring temporal analysis to understand vehicle movement and behavior. It includes detection of:

  1. Over speeding
  2. Red signal violations
  3. Wrong way violations
  4. Prohibited area driving
  5. Obstructing traffic
  6. Reckless driving/one-wheeling
  7. Wrong parking