Efficient Storage Solution for Video Surveillance Archives

This study proposes an Efficient Storage Solution for Video Surveillance Archives by leveraging Anomaly Detection and Video Interpolation techniques. Large volumes of data are stored by traditional video surveillance techniques, which can be expensive to maintain and take a long time to access. The project introduces a modular architecture that integrates advanced technologies, including a MobileNetV2-based anomaly detection model, frame dropping, and video interpolation techniques. In this approach, video streams are captured and subjected to uniform frame dropping to reduce storage requirements while preserving important details during anomalies. The project incorporates a cache management system to store video blocks, prioritizing anomaly blocks for immediate persistence. Persistent storage uses MinIO with a structured folder hierarchy based on date and time, facilitating efficient retrieval and post-event analysis. Additionally, a video interpolation module restores the original frame rate on demand, ensuring smooth video playback without compromising storage optimization. The project aims to bridge the gap between increasing storage demands and the limitations of traditional surveillance approaches, offering an efficient solution for modern video surveillance challenges. The project is evaluated on benchmark datasets on storage efficiency and video metrics like PSNR and SSIM.