Research On Collaborative Coverage And Dynamic Scheduling Of Visual Sensor Networks

With the rapid growth of social security demands,visual surveillance systems which have excellent ability of information capture have attracted more and more attention. Intelligent and collaborative visual sensor networks could be applied to more scenarios. For now, since the cooperation between sensors is insu?cient, most visual surveillance systems in China are ine?cient and unable to fulfill ?exible surveillance requirements. Therefore, research of collaborative scheduling scheme in visual sensor networks has important theoretical significance and practical value for the future video surveillance industry. This thesis mainly focuses on collaborative coverage and dynamic scheduling problems in visual sensor networks. The main contribution of this thesis is concluded below.1. Energy e?cient collaborative coverage algorithm for visual sensor networks.Collaborative coverage problem is concerned in this thesis with the considerations of both area coverage ratio requirement and dynamic target coverage requirement. The so-called Extended Saving Factor Algorithm and Maximal Neighbor Algorithm are presented to satisfy multiple coverage requirements as well as the energy balance among the networks. Network lifetime is thus prolonged by using these algorithms.2. Surveillance strategy design for mobile visual sensor networks based on crowd-sensing task assignment.The problem of coverage over temporary event hotspots in large scenes by crowdsensing is studied in this thesis. crowd-sensing tasks are published and smart phones(mobile visual sensors) users participate in covering temporary event hotspots. To maximize the utility of both supply and demand, a three-level system framework is presented. A task assignment scheme is designed based on this framework. To improve the task execution e?ciency, an improved Lin-Kernighan algorithm is proposed to minimize the task tour by taking consideration of camera’s sensing range. Simulation results show that the strategy presented in this part is effective and reliable.3.Embedded visual sensor network testbed.An embedded visual sensor network testbed is constructed under the requirements of intelligent processing for collaborative coverage and real-time control for scheduling networks. SEED-DEC 6437 DSP platform and EVI D70 C PTZ camera are used to build the visual sensor node. Canny algorithm, Gaussian mixture model and Vibe algorithm are implemented in this embedded visual sensor network testbed. Furthermore, PTZ camera is controlled through RS232 C serial port protocol to guarantee the?exibility. This visual sensor node ensures the possibility of verifying collaborative coverage and dynamic scheduling algorithms of visual sensor networks in practice.