University of Twente Student Theses

Login

Near-Miss Detection on Traffic Intersections with a Distributed Overlapping Multi-Camera System

Weide, Luuk van der (2024) Near-Miss Detection on Traffic Intersections with a Distributed Overlapping Multi-Camera System.

[img] PDF
16MB
Abstract:Intersections are critical areas where a substantial number of traffic accidents occur. Detecting and analyzing near-miss events is key to signaling a high risk of accidents early and taking appropriate measures. Currently, the acquisition of near-miss event data is limited by time and resources. Edge-based cameras offer a new, powerful method for continuously monitoring these events. This study introduces a multi-camera system for near-miss detection. It is designed to operate on a configuration of Technolution's Flowcubes in real-time and on edge. The system leverages rudimentary bounding-box information as input to maintain computational efficiency. It uses a stepwise approach, consisting of middle-point estimation, image-to-world projection, cross-camera association, and trajectory and size estimation. Finally, the resulting shared world-view facilitates the detection and classification of near-miss events. The system has been tested in both synthetic and real-world scenarios, demonstrating robust performance in deriving a cohesive world model, and promising results for near-miss event detection. It presents a scaleable solution for city-wide traffic safety monitoring with minimal infrastructure investment and reduced reliance on human supervision.
Item Type:Essay (Master)
Clients:
Technolution, Gouda, Netherlands
Faculty:EEMCS: Electrical Engineering, Mathematics and Computer Science
Subject:55 traffic technology, transport technology
Programme:Embedded Systems MSc (60331)
Link to this item:https://purl.utwente.nl/essays/98720
Export this item as:BibTeX
EndNote
HTML Citation
Reference Manager

 

Repository Staff Only: item control page