Data reduction in urban traffic sensor network

In a large scale sensor network for traffic surveillance, the data to be transmitted is huge, which leads to high cost of communication. When sensor nodes are connected wirelessly, the situation will be worse. So it´s necessary to reduce the data amount before data packets are transmitted. In this paper, we propose a distributed algorithm base on FCM (fuzzy c-means clustering). Parameters in the algorithm are also discussed. With a set of relative optimal parameters, we perform an experiment. In our experiment based on a traffic flow volume data set, the algorithm shows high-performance with a really high reconstruction precise. The data is compressed evidently before it´s transmitted, saving 46% approximately transmitted.