| Multi-sensor distributed fusion system has become an important means of automatic target tracking in modern war.Multi-source information fusion is the key technology to realize multi-sensor distributed fusion,which can effectively improve the automatic tracking performance of targets and expand the detection range of the system.Compared with other fusion architectures,distributed fusion architecture with center and memory has become a research hotspot in recent years because of its low communication cost and good fusion performance.However,in the realistic multisensor distributed fusion system,the common process noise and the prior information make the complex cross-correlation among the tracks.The clutter interference and multi-target presentation lead to ambiguous track origins.The different time-consuming of local information processing and the communication delay cause the multi-source track to reach the fusion center out of sequence.Those problems finally lead to the significant deterioration of tracking performance.It is a serious challenge to multisensor fusion system.Against the above background,this thesis researches the targetoriented automatic tracking distributed fusion method with center and memory.The motivations and contributions of this thesis are summarized as follows:1.Aiming at the problems of complex cross-correlation among the tracks and ambiguous track origins,this thesis proposes a distributed fusion method with center and memory based on an extended dimension all neighbor association,which approximately achieves the optimal performance of centralized multi-source track fusion.In this method,the extended dimension all neighbor association is used to eliminate ambiguous track origins,and the extended dimension mutual covariance matrix is constructed to remove the complex cross-correlation among the tracks.Simulation results show that the averaged number of confirmed true tracks and tracking accuracy of the proposed method are much better than that of the single sensor automatic tracking,and the optimal fusion performance of centralized multi-source tracking is approximately achieved,but the communication cost is greatly reduced.2.Aiming at the efficient fusion problem of ambiguous track origins and out-ofsequence track of synchronization sensor,this thesis proposes a synchronization sensor out-of-sequence track fusion method based on double prediction decorrelation,which realizes real-time and accurate fusion of multi-source synchronization out-of-sequence track.In this method,all neighbor association technique is used to solve the ambiguous track origins problem of the out-of-sequence track,and the double prediction decorrelation technique is used to fuse out-of-sequence information and update the central track.The simulation results show that,compared with the out-of-sequence discarding method,the averaged number of confirmed true tracks and tracking accuracy of the proposed method are significantly improved,and the fusion performance of the out-ofsequence reprocessing method is approximately achieved,but the fusion time is less and the real-time performance is better.3.Aiming at the optimal fusion problem of asynchronous sensor ambiguous track origins and multi-step out-of-sequence track,this thesis proposes an asynchronous sensor out-of-sequence track fusion method based on equivalent soft association multistep update,which realizes the optimal fusion of multi-source and multi-step out-ofsequence track under the condition of sensor asynchronous.This method solves the problem of the asynchronous and ambiguous track origins of out-of-sequence tracks using integrated equivalent measurement.The central track is iterated to the current time using sequential equivalent measurement.The simulation results show that the averaged number of confirmed true tracks and tracking accuracy of the proposed method achieves the optimal fusion performance of the out-of-sequence reprocessing method,but the real-time fusion is better. |
