Anomaly and Object Detection for Security Surveillance

Human operators’ inability to concentrate on live footage adequately and consistently is a limitation of existing video surveillance systems. After twenty minutes of attention, operators monitoring a single monitor often lose 95% of their focus and judgment. Nowadays, with most surveillance systems having a few dozen cameras, it is obvious that the surveillance duty is beyond the capacity of humans. To increase the effectiveness of video surveillance, this project aims to develop a smart system that will automatically highlight unusual activity and alert the appropriate authorities. The issue is framed as a classification problem, and the objective is to distinguish between typical and unusual conduct in public contexts. Additionally, this project makes use of computer vision and deep learning technology to find numerous things for monitoring. The methods used in this project can broadly be divided into the region- and model-based methods. This has applications in the security and military domains.