Research Of Data Fusion Algorithm Based On Probability Clustering In Wireless Sensor Networks

Wireless sensor network is a network which based on data processing, it have a widly used by applications whether in military or civilian fields. Becouse of its design, its resources and energy are limited. So how to reduce its energy consumption is important. The energy consumption of sensor nodes mainly in the transmit data, that muchlargerthan the energy consumption of calculating and storing, so removing the redundancy and errors of data have great significance.Data fusion is one of effective technologys. In this paper, we discuss cluster protocol based on probability and the data fusion algorithm base on difference vector. As the following:(1) Proposed a state probability prediction model base so as to predicte the distribution probability of collecting data from sensor nodes,supplying a parameter for cluster.Model includes two parameters<S,P>, S is a finite set of states, P is the state transition probability matrix. Using the node calculation and analysis function, we can prediction the probability of data in any moment.The combination of heterogeneous data transformat to the state, generating state probability matrix, and then calculated the balance probabilities, achieve to predict the probability of data.(2) Proposed a Dynamic Cluster Mechanism base on probability of similar. Using the balance probability to calculate the similarity of different sensor nodes, through fuzzy clustering algorithm achieve cluster. Clustering algorithm uses the optimal cost function based on calculus with the iteration method to select cluster head node, determine the cluster head nodes and common nodes by the similarity, using the balance probabilities to calculate sensor’s similarity. This mechanism collect nodes together which have similar balance probability, this can greatly increase the cluster node data correlation, helping improve the fusion rate.(3) Based on the cluster mechanism above, we propsed a data fusion algorithm in cluster head node.Because the high correlation of data in a cluster, we designed the algorithm based on the difference vector, first chooseing a standard value, then record the difference between the data of cluster nodes and standard value. Using this algorithm, redundant data is removed, reducing the amount of network data transmission, while a certain extent, validated the effectiveness of clustering. Finally, the paper simulation results, to analysis and assessment the performance of the algorithm.