A real-life system for identifying and monitoring objects for user-specified scenarios in live CCTV

AbstractThis thesis presents the research and subsequent development of a real life systemcapable of identifying and monitoring objects for user-specified scenarios in live CCTVvideo. More specifically, after a review of the state of the art in both academic methodsand commercial systems, two main novel aspects of the proposed system are detailed.The first deals with the detection of vehicles in static images using a combination offeatures including: corners, lines and colour. The second aspect relates to how motionhas been exploited to detect and track objects within video feeds.The research took place as part of a Knowledge Transfer Partnership, the partnercompany of which had access to many video feeds across a range of differentgeographical locations. This provided the opportunity elicit requirements from the actualend-users and to extensively test and evaluate the system on real data.Results from both video and static analysis systems are demonstrated and evaluatedbefore the Thesis concludes with specific enhancements that could be made to thecurrent system and general recommendations for future work