摄像头网络整体视域优化扩展模型及最优求解

Extended model and optimization for FOV of camera network

研究多摄像头网络环境下监控视域的最大化问题,进而提出全局性优化的粒子群算法获得最佳结果。研究通过调整摄像头的方向,相较于以往的同类系统进一步综合考虑单个摄像头节点的成像清晰度,以及网络的不同监视区域的重要性,从而提出更加适用更加全面的新模型。在这一新的模型下,定义了新的摄像头网络监视优度概念,基于此建立非线性优化方程。为解决这一非线性最大化问题,提出了采用全局优化方法的思路,并给出了以粒子群优化算法求解的具体方案。数值实验表明,该方案可以更好地解决摄像头网络的覆盖优化问题,无论是计算性能还是适用范围,相对于已有的文献和研究报告均有较大的改进。

This paper systematically studied how to maximize the FOV（field of view）of an arbitrary multi-camera network,and proposed to gain the global best results by the particle swarm optimization.The FOV improvement was brought by adjusting all cameras’ orientations automatically.Compared with the conventional researches,further considered the LOD（level of details）of each camera and the importance of the targeted scene.Then got a more realistic and more general FOV model.Under this new model,defined a new concept surveillance goodness,and provided the nonlinear equations for the maximization of the goodness.To solve this problem,suggested the particle swarm optimizer and studied it based on the new model for various scenarios.The experiments show that the proposed model and method work much better than the conventional solutions.Both the performance of the algorithm and the scope of applicable scenarios are largely improved.