GICEDCAM: A Geospatial Internet of Things Framework for Complex Event Detection in Camera Streams

Complex Event Detection (CED) in video streams is increasingly important for surveillance, safety monitoring, and real-time situational awareness. However, detecting complex events remains challenging due to missed object detections, unreliable spatial–temporal relationships, and the high computational cost of existing frameworks. This dissertation addresses these challenges through a manuscript-based thesis consisting of three integrated studies. The first study presents a systematic review of event-matching methods in video-based CED, analyzing 92 papers published from 2012 to 2024. The review shows that Object Detection and Spatio-temporal Matching (ODSM) approaches are the most suitable for near-real-time applications but still suffer from missing simple events, inflexible event-reasoning mechanisms, and limited scalability. These findings motivate the methodological gaps addressed in the remaining two studies. The second study develops an Internet of Smart Cameras (IoSC) architecture that uses edge–cloud collaboration and overlapping camera views to compensate for missed detections. By integrating simple events detected from multiple viewpoints, the IoSC framework significantly reduces false negatives and improves complex-event recognition. Experiments on COVID-19 risk-behavior scenarios show that the IoSC design improves CED accuracy by 1.73× compared to single-camera edge processing. The third study generalizes these insights into GICEDCAM, a geospatial IoT framework that distributes CED workloads across edge, stateless, and stateful layers. GICEDCAM introduces a spatial-event corrector, implemented using Bayesian networks, LSTM models, and trajectory analysis, to reconstruct missing spatial relationships and reduce false positives. Evaluations across four complex-event scenarios demonstrate that GICEDCAM reduces end-to-end latency by up to 36% and lowers computational cost by 45% relative to an open-source baseline, with performance advantages increasing under higher object densities. Among corrector variants, the trajectory-based method offers the best accuracy–latency trade-off for real-time deployments. Together, these three studies contribute a unified, scalable, and geospatially enriched approach to real-time complex event detection. The thesis advances CED research by (1) identifying methodological gaps in the literature, (2) demonstrating how multi-camera fusion reduces missing simple events, and (3) introducing a multi-layer IoT architecture that improves event-level reasoning, scalability, and computational efficiency.

Name thesis:degree_name

Doctor of Philosophy (PhD)

Discipline thesis:degree_discipline

Engineering – Geomatics

Grantor

Schulich School of Engineering

Year dc:date.issued

2026

Author dc:creator

• Honarparvar, Sepehr

Advisors dc:contributor.advisor

• Liang, Steve
• Saeedi, Sara

Committee members dc:contributor.committeemember

• Stefanakis, Emmanuel
• Wang, Xin
• Jabari, Shabnam
• Xu, Lincoln Linlin

• complex event detection
• computer vision
• spatial relationship detection
• cloud computing
• video processing

Statement dc:rights

• Unless otherwise indicated, this material is protected by copyright and has been made available with authorization from the copyright owner. You may use this material in any way that is permitted by the Copyright Act or through licensing that has been assigned to the document. For uses that are not allowable under copyright legislation or licensing, you are required to seek permission.
• Attribution-NonCommercial-NoDerivatives 4.0 International

Licence dc:rights.uri

• http://creativecommons.org/licenses/by-nc-nd/4.0/

Language dc:language.iso

en

DOI

https://dx.doi.org/10.11575/PRISM/51013

OAI identifier oai:identifier

oai:ucalgary.scholaris.ca:1880/123960