Target Detection Based On Data Fusion In Wireless Sensor Network

Wireless sensor networks are applied to all walks of life due to its convenience, selforganization and good performance. As a branch of the field, multi- sensor data fusion has attracted more and more attention; while distributed detection as an important application integration in multi-sensor data fusion has also been favored by many researchers. Target detection based on data fusion in wireless sensor networks has become a hot research of many scholars. This paper puts forward a kind of the topology of the network structure for distributed detection, and analyzes two kinds of detection fusion algorithms. The main work of this paper are as follows:Firstly, combining with the optimal binary tree, this paper presents a new structure based on the traditional tree topology. The structure is a multiple hops network topology, which can improve the reliability of the network. Compared with the parallel topological, our structure can reduce energy consumption in the fusion center; compared with serial topology, it can decrease decision-making reaction time of the system. Different from the traditional tree topology, sensors in our structure are arranged in sequence according to their weights which are proportional to the received signal amplitude. Then the orderly optimal binary tree is generated which combined with the Huffman algorithm. In the structure, the root node receives the strongest signal, and makes a total judgment as fusion center.Secondly, this paper puts forward two kinds of distributed detection algorithms. The first kind of algorithm is dynamic threshold method, namely each sensor makes decisions with dynamic threshold which depends on its child nodes’ decisions. The second kind of algorithm is proportion distribution method, namely, three decision factors associated with sensor decision are allocated with a certain percentage, when the ratio is determined, the sensor will make its own decisions. The results of two algorithms are presented with the form of simulation. Simulation results show that the higher the received signal amplitude of the sensor is, the better the detection performance is. The two algorithms are also compared with parallel structure and serial structures respectively. Under certain conditions, the proposed method is better than the other two algorithms.Finally, this paper describes the principle of the optimal threshold. By introducing the probability of error, we can get a function which is related with the probability of false alarm and the probability of detection. When the probability of error reaches the minimum,the threshold value is the optimal. Comparing sensor measurements and the optimal threshold, the decision-making value can be concluded. We prove that there exists an optimal threshold, and find the optimal value by simulating.