Recent advances in imaging and communication have lead to the development of smart cameras that can operate autonomously and collaboratively to meet various application requirements. Networks of such cameras, called Smart Camera Networks (SCNs), have a range of applications in areas such as monitoring and surveillance, traffic management and health care. The self-configuring nature of cameras, by adjusting their pan, tilt and zoom (PTZ) settings, coupled with wireless connectivity differentiate them substantially from classical multi-camera surveillance networks.

One of the important problems in SCNs is: “How to configure cameras to obtain the best possible coverage of events happening within the area of interest?” As the scale of cameras grows from tens to hundreds of cameras, it is impractical to rely on humans to configure cameras to best track areas of interest. Thus, supporting autonomous configuration of cameras to maximize their collective coverage is a critical requirement in SCNs.

Our research first focuses on a simplified version of the problem, where the field-of-view (FoV) of a camera can be adjusted only by adjusting its pan in discrete manner. Even with such simplifications, solving the problem optimally is NP-hard. Thus, we propose centralized, distributed and semi-centralized heuristics that outperform the state-of-the-art approaches. Furthermore, the semi-centralized approach provides coverage accuracy close to the optimal, while reducing the communication latency by 97% and 74% compared to the centralized and distributed approaches, respectively.

Next, we consider the problem without FoV constraints; we allow FoVs to be adjusted in PTZ dimensions in continuous manner. While, PTZ configurations significantly increase the coverable area, the continuous adjustment nature eliminates any sub-optimality resulting from the discrete settings. However, supporting these features typically results in generating extremely large number of feasible FoVs per camera, out of which only one optimal FoV will be selected. We show that the problem of finding minimum feasible FoVs per camera is NP-hard. However, due to the geometric constraints introduced by the camera's FoV, the problem can be solved in polynomial time. Our proposed algorithm has polynomial-time worst-case complexity of O(n3).


Article metrics loading...

Loading full text...

Full text loading...

This is a required field
Please enter a valid email address
Approval was a Success
Invalid data
An Error Occurred
Approval was partially successful, following selected items could not be processed due to error