Research On Application Of Target Detection In Wireless Sensor Networks Based On Data Fusion

Wireless sensor network (WSN) can be applied in various applications and target detection is one of the most important of them. Sensors sense data in the environment and send them to the sink node after deployed in the region of interest (Rol). The sink node needs to judge the presence of target according to the fusion result. The tremendous data dissemination in network keeps draining the energy of sensor nodes that are energy-constrained. While data fusion can efficiently reduce the redundant data and improve the life time of network. There are two main goals of the research on target detection in WSN:improving the probability of detection and prolonging the lifetime of the network.Firstly, this thesis introduces some basic knowledge and classical algorithms for WSN. The major content is about the challenges of WSN and relation between data fusion and WSN. The relevant simulations and analysis are given in the first section.Secondly, since current algorithms cannot meet the performance requirements, the scan statistics with local vote (SSLV) is proposed in this thesis. Before scan statistics,local sensors perform local vote according to the results of their neighbors and their own.It can effectively increase the probability of detection and decrease the ratio of false alarm.After the local vote, the old expression for global false alarm is not suitable for the SSLV anymore. Hence a new expression has been deduced for the SSLV with the help of Haiman theorem. The operation of local vote also makes the distribution of sensors that have detected the target more concentrated which inspires us to introduce a variable step parameter into the SSLV. In the process of each scan, the scan step is adjusted according to the statistic result. The SSLV with variable step can shorten the scan period without losing any performance. It finds a balance between the continuous scan statistics and the discontinuous scan statistics.Finally, since sensor nodes are energy-constrained, how to reduce the energy consumption is one of eternal topics. As to those protocols based on the fixed sink node,sensors close to the sink are more likely to die because they need to relay data to the sink node frequently. An energy-efficient protocol based on the mobile sink is proposed to balance the overload in the network. The entire network is divided into multiple virtual grids. Every sensor confirms its grid ID according to its location and the grid size. When sensors detect the target, they send the data to the cluster head node in the same grid. The data will transform to the mobile sink by multiple hop among cluster heads. To balance the overhead in the network, the mobile sink node periodically moves to another position according to the left energy and the position of every left cluster head node.