ABSTRACT

We consider an automated surveillance system for public places. In this chapter, by the term public place, we mean a shopping mall, an airport, and so forth. However, with some modification, the present concept of automated surveillance can be applied to any public place like a daily market, railway station, public meeting hall, and so forth. The reason for choosing a shopping mall and an airport as two public places of interest is to handle the surveillance problem at two different levels of surveillance, namely, a bottleneck point (i.e., the entry point of the shopping mall or the airport) and the general area of the shopping mall and airport after the entry point. Based on this particular model of a public place, we design our algorithm for an automated surveillance system. The exits of the shopping mall and the airport are normally kept open and we do not consider any direction of motion at the exit points. At the bottleneck points, the problem of surveillance reduces to an event recognition problem, whereas the problem of surveillance of the areas of a shopping mall and an airport after the entry point reduces to tracking multiple objects. The method of handling event recognition is based on vertical and horizontal histogram analyses of the blob identified by which we can determine the number of objects present in the blob. Subsequently, the detection of a monotonic sequence of centroids of the detected blob is performed. Multiple-object tracking inside the shopping mall and the airport is based on the method of linear assignment and Kalman filter. Using a Kalman filter, we can handle a tracking problem under occlusion.