A novel approach to identify unobvious entities from real time and offline video streaming

Abstract

Security and its related ecosystem have always been given priority in the form of procedures, policies, technology and research. Technology as an assisting component in case of security plays a major role in identifying lapses, loopholes and thus prevent situations to turn into catastrophes. Video surveillance these days has gained significant importance for keeping any place secure. Video cameras are being installed in public places such as malls, theatres, railway stations, super markets, airports and so on. Security personnel monitor these camera feeds from the control centre to observe unobvious entities and manually label the suspected frames. This sometimes turn into lapses due to oversight, fatigue, and negligence because of manual surveillance. This work carried out attempts to overcome these limitations by automating identification of unobvious entities from real time and offline video streams by using the proposed computer vision algorithm. It also proposes to indicate the relative suspicious activity in each frame on a scale of 1 to 10 using the concept of suspectMeter. In addition this algorithm also proposes to reduce the space required for storing suspected frame(s).