Study On Data Fusion Technology Of Wireless Sensor Networks

Wireless sensor network(WSN), combined the technical theory of computer technology, modern communications technology, microelectronics, embedded systems, distributed information processing et al., is an emerging discipline. WSN data fusion can effectively reducing energy consumption and latency of the network while giving users a more realistic and reliable data. Therefore, the study of WSN data fusion has become one of the hot point of WSN research. This paper aim to study and improve the existing clustering routing algorithm and data-level fusion algorithms.This paper is based on the data fusion technology of clustering, the main goal is to collect relevant data within a certain area, and assuming a base station within the network can know the location of each node by positioning technology. Following is the related algorithm:1. Common clustering algorithm (eg Leach algorithm) is prone to the problems of huge clusters and small clusters in the operation, and the node away from the edge of the base station and the cluster is easy to premature death. In this paper, the improved fuzzy C-means clustering algorithm was presented for the problems of Leach algorithm. In the beginning of the network operation, the entire network will be divided into several sub-clusters of equal size by running Fuzzy C-Means clustering on base. Then, the probability that a node elected as the cluster head node was determined based on the membership and the distance from the base station of each node. Because the nodes, at the cluster edges with the low membership and remote from the base station, consumed more energy than other nodes generally. In this paper, the probability of the nodes with low membership and far from the base station elected cluster head node is appropriately reduced to achieve energy balance. In dividing the number of clusters, the concept of the optimal number of cluster head is introduced in order to achieve optimal number of clusters in the network.2. After completion of the network cluster, data fusion can be carried out during transmission. By analyzing the common weighted average fusion algorithm, it can be found that these algorithms are usually build a similarity matrix based on the data collected, such as fuzzy sets, principal component analysis et al.. Then the weight of each sensor can be calculated by take advantage of the similarity matrix. Finally, the integration data obtained using the weighted average. In this paper, the method of constructing weights was improved, using the concept of statistical distribution, taking the collected entire data as a whole and utilizing the T-distribution with a weighted to handle all measurements.3. Since the network is running, the cluster head node may collect exception information, the paper proposes a test optimization solution.Through simulation experiments, Fuzzy C-means clustering algorithm is superior to Leach algorithm in equilibrium of cluster size, the number of survival nodes and the nodes energy balance, etc.. The integration accuracy of fusion algorithm of this paper is better than other methods appeared in other papers.