A New Deep Learning‐Based Video Summarisation Method for Industrial IoT Surveillance Systems
IET Image Processing · 2026-01 · 20 (1)
DOI: 10.1049/ipr2.70324 ↗At a glance
Journal article in IET Image Processing, published online on 2026-03-30. Written by MohammadHossein Ranjbar and 2 co-authors. 57 references. The full text is freely available (open access, CC BY-NC-ND 4.0).
Abstract (original language)
This study introduces an innovative approach to video summarisation, specifically designed for industrial IoT surveillance systems, addressing critical limitations of existing methods, such as inefficiency in dynamic industrial environments, poor adaptability to variable lighting conditions, and suboptimal computational performance. The proposed method integrates MobileNet for feature extraction with a one‐dimensional layer and a Bidirectional Gated Recurrent Unit (BiGRU) enhanced with a self‐attention mechanism to effectively capture long‐range temporal dependencies. A notable innovation of this work is the weighted MSE shot segmentation technique, which dynamically adjusts to varying lighting conditions, enabling precise scene recognition even in challenging industrial scenarios. The proposed model excels in computational efficiency and summarisation accuracy, making it particularly suitable for resource‐constrained environments like industrial IoT. Extensive experiments are conducted on a newly created dataset of industrial videos, specifically curated to reflect real‐world operational challenges, as well as on two widely recognised public datasets, SumMe and TVSum. On the industrial dataset, the model achieved an F1‐score of 0.79, demonstrating its ability to handle structured, repetitive patterns and environmental variability. Additionally, it achieved F1‐scores of 50.8 on SumMe and 63.2 on TVSum, outperforming state‐of‐the‐art methods and showcasing its generalisation capability across diverse video types. By addressing fundamental gaps in existing methodologies and achieving superior performance on both specialised and public datasets, this study establishes a robust foundation for deploying advanced video summarisation techniques. The findings highlight the model's potential to significantly enhance monitoring, decision‐making, and operational efficiency in both industrial and general‐purpose scenarios. Apart from industrial IoT surveillance, the robustness of this method in handling various kinds of videos also highlights it for possible applications in other areas, such as digital media production, where rapid content summarisation can speed up content creation processes, as well as academic research, where efficient summarisation of video datasets is essential for content analysis and understanding.
Cite
APA
Ranjbar, M., Javidan, R., & Shamsinejadbabaki, P. (2026). A New Deep Learning‐Based Video Summarisation Method for Industrial IoT Surveillance Systems. IET Image Processing, 20(1). https://doi.org/10.1049/ipr2.70324
Vancouver
Ranjbar M, Javidan R, Shamsinejadbabaki P. A New Deep Learning‐Based Video Summarisation Method for Industrial IoT Surveillance Systems. IET Image Processing. 2026;20(1). doi:10.1049/ipr2.70324
BibTeX
@article{Ranjbar2026,
author = {Ranjbar, MohammadHossein and Javidan, Reza and Shamsinejadbabaki, Pirooz},
title = {A New Deep Learning‐Based Video Summarisation Method for Industrial IoT Surveillance Systems},
journal = {IET Image Processing},
year = {2026},
volume = {20},
number = {1},
doi = {10.1049/ipr2.70324}
}
EUVIMED