Can Human Visual Surveillance be Improved with Intent Recognition?

In video surveillance applications, trained operators watch a number of screens simultaneously to detect potential security threats. Looking for such events in real time, in multiple videos simultaneously, is cognitively challenging for human operators. This study suggests that there is a significant need to use an automated video analysis system to aid human perception of security events in video surveillance applications. In this paper the performance of humans in observing a simulated environment is studied and quantified. Furthermore, this paper proposes an automated mechanism to detect events before they occur by means of an automated intent recognition system. Upon the detection of a potential event the proposed mechanism communicates the location of such potential threat to the human operator to redirect attention to the areas of interest within the video. Studying the improvements achieved by applying the intent recognition into the simulated video surveillance application in a two phase trial supports the need for an automated event detection approach in improving human video surveillance performance. Moreover, this paper presents a comparison of the performance in video surveillance with and without the aid of the intent recognition mechanism.