Research On Distributed Target Detection In Wireless Sensor Networks

Objective monitoring of the physical world is one of the most important applications of wireless sensor networks. Distributed target detection is to determine the monitored target event occurred or not and it is the premise of all monitoring applications. Under ideal conditions, the judgment of the target event is based on whether the signal value can be determined. However, the actual monitoring environment of wireless sensor networks is complicated. It is often mixed with a lot of random noise which greatly impacts system performance. So how to detect the target successfully under the noisy background is the focus of this thesis.This thesis discusses the problem from four aspects. The first is to analyze the related work in distributed target detection and build the mathematical model; The second is to reevaluate two classic hard decision fusion based distributed detection algorithm with the consideration of energy consumption; The third is based on the sub-cluster structure of the hard decision fusion algorithm and the last is based on serial structure to optimize the routing algorithm for distributed detection. The main work is as follows:(1) The analysis of the two classic hard decision fusion algorithms in energy consumption for the Counting Rule and the Local Vote are the supplements of the results. In the large-scale randomly deployed networks, obtained the result that both in the conditions of the target absence or not, every system energy consumption of the judgment can be approximated only according to its judgment for the total number of 1, and the consumption is linear correlation to the total number. The simulation shows that the following conclusions:The system decision threshold setting is more correlated to false alarm tolerance of each node than that of system. At the same time, although the Local Vote claimed less number of nodes, to sum up the energy consumption of its neighbor's decision-making process of amending is greater than Counting rule. From results of the tolerance of the theoretical and simulation experiments we can see that the approximate formula is accurate.(2) Analyzed the disadvantages of the hard decision fusion algorithm in large-scale, two-dimension wireless sensor networks and given the weighted hard decision fusion model based on clustering structure of the network. This thesis proposes an hard decision fusion algorithm called Weighted based Clustering Decision Fusion Algorithm (W-CDFA). The decisions of sub-cluster upload to their parents'cluster are based on the threshold of themselves. Under the premise of constant false alarm probability, we calculated the decision threshold of each cluster by the Central Limit Theory (CLT). At the same time, in order to narrow the detection area in large-scale networks, we proposed an effective cluster selection strategy called M-CSS based on the membership function. The results of Monte Carlo random experiments show that the W-CDFA algorithm, compared to the Clustering based Counting Rule (C-CR), has certain advantages. W-CDFA algorithm, which is based on the M-CSS, estimates the location of the target event and is able to quickly select valid clusters. Its performance outperforms the W-CDFA.(3) The source node of distributed detection required a routing mechanism to transmit their decisions to the fusion center. We analyze the limitations of hard decision fusion in serial-based network structure. Based on the value fusion, we converted the target detection performance into a weight of the routine link in according to find one path which detection probability is maximum under constant false alarm rate. Due to the topology of wireless sensor network dynamically changes and target detection applications often require real-time correspondence, we proposed a multi-path routing algorithm DD-ACMRA based on the improved ant colony algorithm for distributed target detection. Simulation results show that DD-ACMRA has a preferable performance in target detection, latency and network energy consumption according to a specific demand.