Research On Distributed Decision Fusion Algorithm Under Imperfect Channels

Sensor network target detection refers to deciding whether the target is present with the help of sensor networks in a detection area.In recent years,sensor network target detection has been extensively studied,and practically applied in many fields such as military,agriculture,and marine resource detection.Usually,most sensor nodes in a sensor network are powered by batteries in practical application.When the detection area becomes more complex,replacing the battery for sensor nodes that run out of power would be difficult,complex and inefficient.In addition,for a sensor network target detection system,the system's detection probability is always used to measure its detection performance.Therefore,how to reduce the energy consumption,improve the system's lifetime,and maximize the detection probability of the system are the main concerns and research points of this thesis.Considering that practical application is the ultimate purpose of theoretical research,therefore,a sensor network target detection model with imperfect channels is established in this thesis.From the perspective of energy consumption and detection performance,this thesis investigates sensor network target detection algorithms to reduce the system's energy consumption and improve the system's detection performance.Firstly,this thesis introduces the basic network structure and characteristics of the sensor network,three common data fusion models and their corresponding structures,common data fusion methods,basic theories of target detection,and clarifies the difficulties and challenges of sensor network target detection based on data fusion.Then,based on the distributed detection fusion system model under non-ideal channels,this paper proposes two target detection fusion algorithms to balance the system's energy consumption and detection performance effectively.In the distributed data fusion,soft decision fusion and hard decision fusion are two common fusion methods.Compared with soft decision fusion,hard decision fusion has worse detection performance but consumes less energy consumption.In order to make full advantage of soft decision fusion and hard decision fusion,this thesis proposes a soft-hard combination decision fusion algorithm based on network clustering.The proposed algorithm could reduce the system's energy consumption from the perspective of reducing the amount of transmitted data and shortening the data transmission distance.At the same time,clustering could share the system load to each sensor node efficiently,which can improve the system's lifetime.In addition,soft decision fusion within clusters and hard decision fusion between clusters can effectively guarantee the detection performance of the detection system.However,no matter in soft decision fusion or hard decision fusion,the number of bits transmitted by each sensor node is always same and would not be changed.Considering the differences of each sensor node in the environment,distances from the fusion center,credibility of local decisions,and remaining energy and so on,this thesis proposes an optimal dynamic bit allocation algorithm,in which the number of bits allocated to each sensor node could be different.This algorithm transforms the problem of finding the optimal bit allocation scheme into multi-objective optimization problem to maximize the system's detection probability and minimize the system's energy consumption.