Data Fusion System Design And Implementaiton Based On Parallel Computing Techniques

In recent years, with rapid development of the "Internet of Things"technology, the data fusion technology has penetrated into all aspects of human life. With the increase of sensor type and sensor performance, the traditional data fusion system can't meet the needs of scalability, data processing level and monitoring accuracy in the field of water target public opinion monitoring. For these reasons, this paper presents a data fusion system architecture scheme based on parallel computing technology for vessel targets.The data fusion system based on parallel computing technology, based on heterogeneous sensor network and protocol analysis system, forms a distributed real-time data analysis platform between the physical layer and the application layer.Firstly, this paper analyzes the data characteristics of the data fusion scene. Then combined with the Lambda architecture concept, this paper design a three-layer system architecture which contains real-time computing, off-line computing and computing result output. The real-time computing part is mainly responsible for multi-source trace correlation and situation estimation. Off-line computing is mainly used for data statistics of threat characteristics, and data persistence storage. And computing result output part is mainly used for caching the data. Secondly, some key algorithms, such as time registration algorithm and trace correlation algorithm, are improved and optimized in the process of data fusion, so that they can run better in the parallel computing framework. Finally, based on the data fusion system designed and implemented, this paper does a number of experiments about the algorithm simulation and system performance. The experimental results show that the time registration algorithm and trace correlation algorithm can meet the requirements of accuracy, real-time and system can meet the expected requirement in both reliability and computational performance.