Region and corners based feature extraction techniques with intersection estimation for moving object behaviour analysis

Abstract

Event detection and abnormal behaviour analysis are the two major newlineconcerns in many of the video-surveillance applications. The process of newlineactivity recognition has been well concentrated in underlying vision newlineintelligence platform which can be deployed on various fields like health care newlinemonitoring, traffic surveillance, infant monitoring and defence security newlinesystems. The research work on event behaviour analysis in real-time newlinevideo-surveillance is quite sparse and is considered to be a challenging task newlinewith respect to illumination variation, occlusion, viewpoint variation and newlinetarget orientation. The target behaviour has to be monitored continuously by newlinefixing either static or dynamic cameras for taking impulsive action towards newlineprotecting and providing security to the civilians in public places. newlineThe video analytic techniques or forensic methods on human newlineobservations are carried out to classify the normal and abnormal behaviour. newlineMoreover, an unusual event (vehicle-pedestrian collision or human-human newlineinteraction) is hard to distinguish as they need to be characterized using newlinehigh-level interpretation in the video sequences. A proactive vision-based newlineevent behaviour analysis technique has been proposed to detect multiple newlinetargets for understanding the unusual behaviour of a pedestrian which newlineinstantly alerts the individual before the incident could happen. newline newline