To meet the need of large scales plates and the real-time performance of DataFusion system , a stellated master-slave parallel computing pattern was used in thispaper after some analysis about the principle and process of deducing The computingtarget was divided according to moderate granularity and dispatched by threadmechanism. A parallel virtual machine was constructed by PVM, and a method based onCOW, which runs successfully on NCI Cluster 2000,sloves the problem of real-timebasically. The results are good. An algorithm of attribute/track correlation was presented inData Fusion system mainly using high accurate attribute parameter. Many parameters wereused to eliminate incorrect result and prove the correctness of system...
|