Deep learning for secure imaging and video surveillance in smart cities: from bibliometric mapping to real-world implementations

作者：Samad Olalekan Olansile, Oluwatosin Ahmed Amodu, Mohamed Sani Adam, Isaac Oluwafemi Elesemoyo, Raja Azlina Raja Mahmood · 发表于：Frontiers in Imaging · 年份：2026 · DOI：10.3389/fimag.2026.1854187 · 研究领域：Adversarial Robustness in Machine Learning、Advanced Neural Network Applications、UAV Applications and Optimization

The integration of imaging sensors, unmanned aerial vehicles (UAVs), and satellite platforms has expanded capabilities in surveillance, smart-city systems, and remote sensing. Advances in deep learning and computer vision enable automated detection, tracking, and scene understanding across diverse environments, including safety-critical settings; however, these systems impose requirements for security, privacy, and adversarial robustness. This paper presents a dual-database bibliometric and science-mapping analysis of deep learning research for secure imaging and video surveillance. Two independently constructed corpora, derived from Scopus (𝒟 S ) and Web of Science (𝒟 W ) for the period 2010–early 2026, are analyzed using VOSviewer-based keyword co-occurrence clustering to identify dominant research themes and methodological trends across surveillance, smart cities, remote sensing, and computer vision; annual publication trend interpretation is restricted to complete calendar years from 2010–2025. Quantitative bibliometric analyses (e.g., publication trends, geographic distribution, institutional productivity, and keyword co-occurrence structure) are conducted on both 𝒟 S and 𝒟 W , with cross-database comparison used for validation. Complementarily, detailed analysis of prominent keywords and inter-cluster connectivity, as well as the review of real-world deployment systems and field-dataset studies, are conducted based on data from the Scopus dataset (𝒟 S ), which provides ...