The Research And Implementation Of Multiple Sensor Data Fusion Technology

Data fusion technology has attracted much attention in recent decades,and its application has been extended to all aspects of the life of the residents.In the monitoring system of the Internet of Things,we can get the detailed description information of the target by detecting the characteristic data of different angles of the unknown target through a variety of heterogeneous sensors.Data fusion technology is the key to use the information effectively and reasonably.At present,fruitful achievements have been made in the research of theory and algorithm.In principle,the greater number of sensors is introduced,the more fully understand the target and the environment.But with the increase of the number of sensors,the amount of data need to process by the system increases dramatically.It brings great challenge to the traditional single machine data fusion system,and the best way to solve this problem is to change the structure of the traditional system by using the method of distributed parallel processing.Whereas,it is necessary to design and implement a multi-sensor data fusion system based on Storm.Storm is a kind of very popular distributed computing engine,it is more suitable for data fusion scenarios than Hadoop.Storm designs a specific set of primitives for processing real-time messages and for the operation of the database.Storm can easily and quickly carry out complex calculation in the computer cluster,and ensure that every message in the system can be processed in time.This paper analyzes and designs the data fusion algorithm,then applies storm flow calculation engine to multi-sensor data fusion system,then implements and operates each module of the data fusion in the framework of the Storm.This approach solves the performance bottleneck problem of centralized processing.This article firstly introduces the research background and related technologies,and expounds the development trend and existing problems of the data fusion,and analysis the feasibility of applying Storm to the data fusion system.Then it uses the data stream model to represent the original observation data from various sensors and divides the data stream using sliding window mechanism,which is convenient for each module of the system to process the data.Then,it proposes time registration,track association and track fusion algorithm on the basis of studying the classical algorithm of data fusion.Then it achieves the various modules of the system in the Storm framework.Finally,it tests the validity of related algorithm and the performance of the system.