Intelligent Video Surveillance for Human Behavior Analysis and Anomalous Activity Detection using Deep Learning

Human Behaviour Monitoring and Categorization using Deep Learning Techniques has the potential to strengthen public safety in sensitive and public locations by enabling the monitoring of human activities and identifying anomalous behaviours. Many crimes, such as terrorism, theft, and vandalism, can be avoided by keeping an eye on human activity in sensitive and public places including bus stops, train stations, airports, banks, shopping centres, schools, and colleges. Over the past decade, the field of visual surveillance has experienced a notable surge in publications dedicated to abnormal activity detection. Manual video monitoring, being resource-intensive and inefficient for continuous oversight of public spaces, is surpassed by intelligent video surveillance systems. These systems leverage sophisticated deep learning algorithms, notably Convolutional Neural Networks (CNN-2D) and the VGG-19 architecture, operating in real-time to recognize and categorize human activities as normal or unusual, subsequently triggering timely alerts. The presented study introduces an intelligent video surveillance system that exploits CNN-2D and VGG-19 for feature extraction and activity classification. Evaluation of a diverse dataset underscores its compelling results, demonstrating high accuracy in recognizing suspicious activities and outperforming traditional methods. The comparative analysis of deep learning techniques, including CNN-2D and VGG-19, underscores the superiority of this approach in terms of accuracy, efficiency, and robustness.