Fuzzy-based path analysis

Video surveillance can be a very powerful tool in the fight against crime, by accurately monitoring human activities. Nevertheless, most surveillance systems today provide only a passive form of site monitoring. Extensive video records may be kept to help find the instigator of criminal activities after the crime has been committed but preventive measures usually require human involvement. In addition to this, there is a need for large amounts of data storage to keep up to several terabytes of video streams that may be needed for later analysis. In order to achieve any form of real-time monitoring, guards often need to be employed to watch video feeds for hours on end to recognize suspicious, dangerous or potentially harmful situations. In a multi-camera scene monitoring system, this can be quite infeasible as there can be up to 20 to 50 cameras on average in a large building complex such as an airport or shopping malls. Intelligent video surveillance aims to reduce or even eliminate the need for human supervision of video feeds, and continuous recording. Having such a system will provide numerous other facilities and services to operators and emergency teams, by conducting behavioral analysis on incoming video feeds and detecting unusual or suspicious behavior. Behavioral analysis itself can be applied to numerous features extracted from video sequences including path detection and classification of which several methods are reviewed here. In this paper, we investigated a fuzzy inference engine approach to identify the human trajectories based on the paths that had been modeled by a self-learning system.