Human Security: Intelligent Video Surveillance for Criminal Activity Detection

Abstract

As cities keep expanding, ensuring public safety will become increasingly difficult considering the complexity and unexpectedness of crime, the density, and dynamism of crowds. Traditional public safety surveillance technologies have been subject to high false alarm rates and lack contextual awareness, which makes monitoring inefficient and reactions slower. The evolution of AI in modern computing, especially video processing through deep learning, opens new directions for building adaptive and efficient crime-detection systems that will build capabilities of threat identification and behaviour analysis in real time. This review of relevant academic research underlines state-of-the-art work on deep learning and surveillance systems, with convincing evidence for their ability to reduce false alarms and improve accuracy in real public settings. The review points to the scalability and interpretability of AI systems as key to supporting security personnel, communities, and society with reliable and automated decision-making. The review discusses how contextual interpretation, body pose estimation, behavioural modelling, and multilayered threat detection advance public safety. Eventually, this work introduces a novel intelligent surveillance architecture capable of applying deep learning techniques, as mentioned above, to establish a reliable crime detection system for urban capacity. The larger goal is to adapt best-in-class safety infrastructure with next-generation AI capability and deploy these into practice for decision-making reliability.