INTELLIGENT VISUAL SURVEILLANCE

In this book, we describe the proposed pedestrian §classification and tracking system that is§able to track and label multiple people in an §outdoor environment such as a railway station. We§propose an approach that combines blob matching with §particle filtering to track multiple people§in the scene, i.e., the proposed method selects the §successful features of blob matching and particle§filtering for tracking. In addition, a§novel appearance model derived from the colour §information from both the moving regions and §theoriginal input colour image is proposed to track §people in the event of poor foreground extraction.§Additionally, the proposed appearance model also §includes spatial information of the human§body in both vertical and horizontal directions, §making location more accurate. Furthermore, a novel §method to extract cars from moving regions including §shadow area, based§on shape and colour information, is proposed. §Chamfer template matching score, and non-shadow§region edge score, are applied as the shape §information; while the shadow confidence score§(SCS) is used as the colour information.