Research On Parallel Computing In Target Identification

To meet the needs of large-scale platforms in battlefield and the requirement of real-time performance of the Data Fusion system, the parallel computing with MPI in the target identification of Data Fusion system is discussed in this thesis. And this thesis deals with the load balancing in parallel target identification of the Information Reconnaissance and Data Fusion Systems particularly. Being analyzed, it has been found that the particularity of load balancing in this case is the requirement of history identification data in the Dempster-Shafer Evidence Fusion. This limits the free dispatching of tracks. This thesis provides an improved dynamic load balancing algorithm and presents the implementation. The algorithm considers the requirement of history identification data, avoids the risk of parallel identification degenerating into serial actually. And at last, we test the parallel target recognition module online. The result shows that the bottleneck of data fusion system is resolved and the goal of our system is reached.