Intelligent video analytics for real-time occupational safety management on construction sites
Year: 2026
Type: article
Abstract: Real-time identification of construction workers and compliance verification under dynamic drone surveillance face severe bottlenecks, primarily due to the vulnerability of standard QR codes to motion blur, acute viewing angles, and low resolution under full-coverage personal protective equipment (PPE). To address this gap, this study proposes an optimized, detector-free video processing pipeline featuring independent non-blocking stages, RTSP/WebRTC streaming, and Redis-backed temporal deduplication to evaluate fiducial AprilTags versus QR codes across varying badge heights and tracking conditions. Empirical findings demonstrate that AprilTag achieves a superior decode yield of 0.623 in detector-free configurations. It outperforms QR codes (0.161) by 3.9× and maintains operational viability at minimum badge heights of 24–32 pixels, whereas QR codes fail below 48 pixels. Furthermore, track-level temporal accumulation reveals that AprilTag requires only 4 sightings to attain 95% worker identification confidence, compared to 18 sightings for QR codes, while ensuring low-latency streaming and efficient storage control. These results establish robust camera placement specifications and track-level accumulation as core metrics for real-industrial video surveillance. Future research will investigate multi-drone orchestration and adaptive image enhancement techniques.
Authors: Abdulla Omarov(Conceptualization, Data curation, Formal analysis, Funding acquisition, Investigation, Methodology, Resources, Supervision, Visualization, Writing – original draft, Writing – review & editing), Askar Zhussupbekov, Sungat Akhazhanov(Data curation, Formal analysis, Validation), Victor Kaliakin(Writing – original draft, Writing – review & editing), Olzhas Bibalayev(Conceptualization, Formal analysis, Software, Validation, Visualization), Zhibek Zhamaladinova(Data curation, Validation), Daniel Mikhailov(Data curation), Ainur Montayeva, Gulzhanat Tanyrbergenova(Formal analysis), Ankit Garg(Writing – original draft)
Language: English
Subfield: Information Systems
Field: Computer Science
Domain: Physical Sciences
Open Access status: diamond
OpenAlex ID: W7234010094
