Optimizing Public Security Through Intelligent Video Object Detection and Super‐Resolution With SESPCN and Deep Fusion Technique

Video surveillance is the main requirement of smart urban development and government security, and the traditional systems need live monitoring with manual supervision of the information; hence, there is a limitation on the practical use of the data gathered through surveillance. In the study, the deep fusion network architecture is considered the main architecture to identify objects in video materials, followed by enhancing them for real‐time security systems. The system employs you only look once v9 (YOLOv9) to detect objects precisely, thus allowing fast tracking and identification of suspicious items. The framework uses adaptive motion analysis to choose significant frames for efficient analysis purposes and automatic event monitoring procedures. Semantic efficient sub‐pixel convolutional neural networks (ESPCN) perform high‐resolution enhancement using low computational resources, which feature pyramid networks (FPN) systems use to extract multi‐scale features for better image clarity during super‐resolution processing. The Poisson Image Editing method allows users to combine. To maintain visual coherence while incorporating the enhanced items into video footage, use File Writer. This paper introduces a novel approach that combines segmentation models, detection models, resolution enhancement tools, and a post‐processing function. This approach enhanced real‐time video surveillance with improved accuracy and processing speed. The experimental data confirms the system works effectively in real‐life surveillance through a 90.5% accuracy rate achieved alongside 50 predictions per second processing abilities.