Simultaneous Boundary Partitioning and Cameras Synchronization for Optimal Video Surveillance

Abstract

This paper proposes a real-time distributed algorithm for a team of smart cameras to self-organize and perform video surveillance of an open boundary. In particular, our algorithm simultaneously partitions the boundary among the cameras, and synchronizes the motion of the cameras to optimize the surveillance performance. We focus on the detection of smart intruders, who are aware of the cameras configuration at each time instant, and who schedule their motion to avoid detection for as long as possible. We consider both the worst-case and the average detection times of smart intruders. Our algorithm achieves minimum worst-case detection time, and, under some reasonable assumptions, constant-factor optimal average detection time.

Camera Network Coordination for Intruder Detection

“…If the patrolling windows are not given but are still required to be a partition of Γ, then the longest cameras sweeping time, and hence the worst-case detection performance, can be minimized by solving a min-max partitioning problem [7], [18], [14]. We will discuss this partitioning problem in Section V, where we develop an algorithm to simultaneously partition the boundary and coordinate the motion of the cameras.…”

“…Complementary approaches based on numerical analysis and game theory for the surveillance of two dimensional environments are discussed in [11], [12]. Finally, a preliminary version of this work was presented in [13], [14].…”

Multi-Agent Autonomous Surveillance: A Framework Based on Stochastic Reachability and Hierarchical Task Allocation

“…Camera surveillance problems can also be viewed in the context of pursuit-evasion games [20][21][22][23] where the evader actively tries to avoid detection by the pursuer. Even though surveillance methods based on game theory are very interesting, they can result in nonrobust behavior whenever the evader strategy or the environment differ from the assumed one (see Ref.…”