Object detection for construction site safety monitoring based on Yolov8 model

Read full text Summarise with VIRAT
Authors: Mingyue Qu, Jingjing Zheng
Cited by: 0
References: 24

Summary

No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.

Summarise with VIRAT Free to try with your Science Catalogs account.

Abstract

Accidents frequently occur at construction sites, and traditional video surveillance relies on manual inspections, which have problems of delayed response and high false alarm rate.This paper addresses the challenge of precise identification of small and occluded targets in complex environments and proposes a real-time detection framework based on the improved you only look once version 8.By integrating multi-scale features and optimising the network structure, the system achieves an average precision of 92.3% on the public safety helmet dataset, which is 8.5 percentage points higher than the benchmark model you only look once version v5; the processing speed on edge devices reaches 45 frames per second, meeting the requirements of real-time monitoring.This method enables automatic identification and warning of behaviours such as wearing safety equipment and intrusion into dangerous areas, providing effective technical support for building an intelligent safety defence line.

Fields of Study & Subjects

Advanced Neural Network Applications Video Surveillance and Tracking Methods Infrastructure Maintenance and Monitoring

Publication details

Journal / source
International Journal of Reasoning-based Intelligent Systems
Volume
1
Issue
1
Pages
66–82
Publication date
1 January 2026
Language
EN