Abstract: Surveillance video anomaly detection is vital for enhancing security by identifying unusual activities in video footage. Traditional methods often face challenges with high false alarm rates and scalability. Recent advancements in deep learning, including convolutional neural networks (CNNs) and autoencoders, have improved anomaly detection by analyzing patterns in video data. This paper reviews various approaches to anomaly detection, such as motion analysis and deep learning techniques, while addressing challenges like real-time processing, data imbalance, and the need for large labelled datasets.
Finally, we discuss future directions, including multi-modal data integration and more efficient models for diverse surveillance environments.

Keywords: Real-time surveillance, Anomaly detection, Machine learning, Deep learning, Convolutional Neural Networks

Downloads: | DOI: 10.17148/IARJSET.2025.1211021