Coverage management for mobile targets in visual sensor networks

Visual sensor networks (VSNs) are becoming increasingly popular in a number of application domains. A critical ability of such networks is to self-configure to minimize the need for operator control, improve scalability, and reduce cost. One of the areas of self-configuration is camera coverage control: how cameras adjust their field-of-view to allow automatic tracking of maximum number of targets. The problem is shown to be NP-hard for stationary targets, and efficient centralized, distributed and semi-centralized heuristics exist that perform close to optimal. For stationary targets camera configuration is a one-time activity and can happen offline and before the actual deployment. In contrast, if the targets are mobile, as the targets move away from their recorded positions, the cameras need to configure dynamically and in real-time to ensure coverage accuracy. In this paper, we propose several policies for automatic control of the cameras with a goal of coverage maximization for mobile targets. We study these policies using important performance metrics such as coverage gain, adaptability, scalability, and energy consumption. Our results indicate that factors such as target mobility models, target and camera scales/densities, and target velocities have significant impact on the performance of a given policy. For most of the scenarios, we found that the protocols that take into account non-local information (e.g. neighborhood information) and have self-adapting parameters (e.g. frequency of camera configurations) outperform the protocols that are either purely local or purely global and have non-adaptive parameters.