Pedestrian Crosswalk Inspection System

About Us

KTUTrafik, formally titled Pedestrian Crosswalk Inspection System Operating with Image Processing Technology, is a graduation project developed at Karadeniz Technical University by Barış Vatan USTA, Fikret TELLİ and Mehmet Okyay SÜTIRMAK under the supervision of Asst. Prof. Mehmet ÖZTÜRK. The realization of this system was made possible through the financial support of the TÜBİTAK 2209-A Research Project Support Programme for Undergraduate Students. Furthermore, the project's objectives are purposefully aligned with the United Nations Sustainable Development Goals (SDGs) 3, 9, and 11.

Protecting Pedestrians Through Intelligent Monitoring

The Pedestrian Crosswalk Inspection System is an intelligent traffic monitoring platform designed to make urban crossings safer. It automatically detects vehicles that fail to yield to pedestrians at crosswalks, documents each violation with photographic evidence, and makes the records accessible through a secure web interface. Our goal is to support safer streets by turning everyday camera footage into reliable, actionable safety data.

How It Works

Edge Detection. A camera connected to a Raspberry Pi 5 with a Hailo-8L AI accelerator runs YOLO-based computer vision in real time, detecting the crosswalk, pedestrians, and vehicles directly on the device.

Violation Analysis. When a pedestrian is actively on the crosswalk and a vehicle enters the same area, the system identifies it as a violation and captures the moment as photographic evidence.

Cloud Relay. The captured evidence and the live video stream are sent to a cloud server, which distributes the stream to all connected viewers without overloading the edge device.

Plate Recognition. Each violation image is processed to read the vehicle's license plate, which is then linked to the record so it can be looked up later.

Public Access. Citizens can check a plate for violations, view live footage, and explore traffic safety statistics through this website.

Real-Time AI

On-device YOLO inference accelerated by the Hailo-8L delivers fast, continuous detection rather than delayed snapshots.

Evidence-Based

Every violation is documented with a timestamped image, so records are verifiable and transparent.

Live Streaming

A relay architecture streams the monitored crossing to any number of viewers with minimal load on the camera.

Secure by Design

Encrypted connections, authenticated devices, and a protected admin panel keep the system and its data safe.

Technology

The system is built on a combination of edge computing, computer vision, and cloud infrastructure.

Raspberry Pi 5 Hailo-8L AI Accelerator YOLOv8 Optical Flow Tracking Flask + Socket.IO License Plate Recognition Cloud Relay