Transfer Learning based Video Surveillance for Abnormal Activity Detection

With the prolific rise in the number of surveillance cameras in both inner and out-of-door locales, there’s an increasing need for an intelligent system that detects abnormal or unusual events. Surveillance of CCTV videotape recordings manually is a tiresome process where the response to a detector event must be instantaneous. Unusual exertion discovery plays an important part in surveillance. This study plans to develop a novel system to automate the task of videotape surveillance and identify the unusual activities similar to road accident discovery, fall discovery, fire discovery, and armament discovery. The fall discovery detects colorful posture-grounded events in a typical senior monitoring operation in a home surveillance system. fire discovery is a basic and efficient algorithm which in the event of any fire disaster automatically detects in the surveillance area via real-time analysis of the recorded video by the system. A fast and precise discovery process for early fire advising can be developed by observing and exercising fire event features. Auto crash discovery helps in recycling the surveillance videos that are being recorded and can return back directly to emergency services without any intervention. This makes it time efficient and reduces the latency in communication as every second is important to save lives during an accident Weapon discovery is one of the ways to reduce violence.