Dynamic Network Flow Optimization for Task Scheduling in PTZ Camera Surveillance Systems

This paper presents a novel approach for optimizing the scheduling and control of Pan-Tilt-Zoom (PTZ) cameras in dynamic surveillance environments. The proposed method integrates Kalman filters for motion prediction with a dynamic network flow model to enhance real-time video capture efficiency. By assigning Kalman filters to tracked objects, the system predicts future locations, enabling precise scheduling of camera tasks. This prediction-driven approach is formulated as a network flow optimization, ensuring scalability and adaptability to various surveillance scenarios. To further reduce redundant monitoring, we also incorporate group-tracking nodes, allowing multiple objects to be captured within a single camera focus when appropriate. In addition, a value-based system is introduced to prioritize camera actions, focusing on the timely capture of critical events. By adjusting the decay rates of these values over time, the system ensures prompt responses to tasks with imminent deadlines. Extensive simulations demonstrate that this approach improves coverage, reduces average wait times, and minimizes missed events compared to traditional master-slave camera systems. Overall, our method significantly enhances the efficiency, scalability, and effectiveness of surveillance systems, particularly in dynamic and crowded environments.

💡 Research Summary

The paper proposes a comprehensive framework for real‑time scheduling and control of Pan‑Tilt‑Zoom (PTZ) cameras in highly dynamic surveillance environments. The core idea is to fuse accurate motion prediction, obtained via a Kalman filter for each tracked object, with a dynamic network‑flow optimization that decides which camera should observe which location at each discrete time step.

Prediction Layer – Each object is equipped with an independent Kalman filter that estimates its future position based on current state and a linear motion model. The filter outputs a predicted location for the next planning horizon, providing a reliable “track node” (K​j,t) that the scheduler can use. Prediction errors in the authors’ simulations stay below 0.3 m on average, which is sufficient for precise PTZ aiming.

Dynamic Network‑Flow Model – The surveillance system is represented as a time‑expanded directed graph G = (V,E). Nodes consist of (i) camera supply nodes R​i,t (one per camera per time step, supply = 1), (ii) high‑resolution track nodes K​j,t, (iii) low‑resolution fixed‑area nodes F​j,t, and (iv) periodic demand nodes D​j,τ that enforce a minimum visit frequency T for each fixed region. Arcs connect cameras to either track or fixed nodes, and also link successive time copies of the same track or fixed node to preserve flow continuity. Each arc carries a binary flow variable x, indicating whether a camera is assigned to that location at that time.

The objective maximizes the total assignment value:

\