Figure 1

Overall architecture. Top down: a single interface is used for direct control of any server/camera and high-level processing (e.g., user-defined face recognition). The server connects to a switch which hosts a database and joins two sets of servers: a series of mid-level (e.g., feature extraction) and low-level processors (e.g., detecting moving objects). The switch connects to routers which communicate with wireless bridges connected to the IP cameras.