Pedestrian behaviour analysis using the microsoft kinect

An important challenge for brick-and-mortar retail businesses is how to monitor the interest of customers in displays and products in a store. We investigate the effectiveness of the Microsoft Kinect as a sensor to monitor the behaviour of pedestrians, which may reflect their interest in a store display or advertisement. In a controlled environment, participants acting as pedestrians are requested to show different levels of interest as they pass an object that is being monitored by a Kinect sensor. The sensor collected measurements such as the positions and orientation of the pedestrian's body and head, which were analysed to detect the movements of the pedestrians and thus infer the behaviour they exhibited. Our results demonstrate that when the Kinect is able to detect a behaviour that indicates a pedestrian's interest in an object, then the classification of the level of interest in terms of the type of behaviour is reasonably accurate under varying light conditions and numbers of pedestrians.