Research On The Key Technology Of Information Fusion In Battlefield Environment

Information fusion technology is a new cross subject,and the military demand is the source of the development of information fusion technology though it is booming in the application of various industries and fields.The aim is to realize the higher unification of the physical domain,information domain and cognitive domain for the complex battlefield environment by the information fusion technology.After more than forty years of development,there are imperfections in the theoretical system and algorithms although the information fusion technology in the military application has made considerable progress.In the dissertation,the key technologies of information fusion technology in battlefield environment are studied systematically,and the main contributions are as follows:1.The relationships between the function and the information flow are not clear,and also the functional relationships between the situation assessment and the threat assessment are confusion when assessing the battlefield situation,a structure model of situation assessment is proposed to settle the problems.The structure model can construct a kind of the hierarchical information relationships between the awareness situation、the perception situation and the prediction situation,which can make sure the information relationships between the functional modules in situation assessment are more clear,and the functional differences between the situation assessment and the threat assessment are more unequivocal and easily applied to engineering.2.In view the shortcomings of the large amounts of computing and the slow convergence for the data association,a swarm intelligence algorithm is proposed to solve the problem.Firstly,the problem of multiple targets data association is transformed into the problem of constrained optimization.Then,with the improved PSO algorithm,which has fast convergence in the start search optimum solution,and with the improved ACO algorithm,which has fast convergence in the last search optimum solution,the proposed algorithm combine the advantages of the two intelligent algorithms,and achieved the aim of the fast correlation.At last,by comparing with the JPDA(Joint Probabilistic Data Association)algorithm,the results shows that the two algorithms almost have the same accurate,but the time consumption of the proposed algorithm decrease from more than 20 seconds to about100 milliseconds.By comparing with the ACDA(Ant Colony Data Association)algorithm,the proposed algorithm improved about 7% in association accuracy,and the time consumption reduced about 50%.3.It has no effective method to achieve the target grouping result when the two sides struggle in the state of confrontation,a new algorithm is put forward to solve the problem of targets clustering in multiple formations combat condition.When grouping the enemy’s targets,the improved Chameleon clustering algorithm is used to form the space groups firstly,and then calculating the advantage function of the attack elements and the attack matrix,the interaction group is achieved by analysis the attack matrix at last.When grouping our targets,the relative attack superiority matrix is calculated firstly,then transform the offensive combination question into the optimization problem,and the improved genetic algorithm is used to obtain the interaction group at last.Compared the referenced algorithm with the proposed algorithm by simulation,the results show that the two algorithms are consistent in space grouping,but the proposed algorithm has the higher accurate in interaction grouping,and the proposed algorithm can reach twenty milliseconds in real-time performance.4.There are some shortcomings for the intensive group target tracking algorithms:filtering model parameters are difficult to adapt to the needs of the environment,the extended state estimation of the group target can not reflect the extended state’s changes effectively.For the problems above,an parameters adaptive tracking algorithm is proposed for the dense group targets tracking in the Bayesian filtering framework.The transition probability of the model state transition matrix is modified by the posterior information to adjust the parameters adaptively.Then,by automatically prejudge the splitting and merging of the group targets,as well as correct the extended state sensitivity online,it can improve the estimation performance of the extended group target state.At last,compared with the referenced algorithm,the error of the proposed algorithm decrease by about half in distance and speed,the split group target can be detected only 3 cycles after it splitting,and the new group targets can be tracked continuously and effectively.5.The characteristics of the battlefield environment are as follows: situation elements changes continuously,information representation and reasoning is certainty or uncertainty,the process of the inference and estimation are real-time.In order to meet the needs above,a situation assessment algorithm is put forward to achieve the online learning parameters by reasoning the Dynamic Bayesian networks.The forward recursive method is used to reason the DBN,in view of the phenomenon of the small sample properties for the situation elements,the Dirichlet distribution is selected as the prior distribution to estimate the network parameters in order to improve the performances of the precision and the real-time.Compared with the EM algorithm,the proposed algorithm can decrease by 6-7 times in average and maximum time-consuming,the biggest threat degree estimation probability increased from0.58 to 0.71 for the hidden variables,the estimation accuracy of parameters A and B increased significantly,and the accuracy of the parameter B improve faster with the increase of the parameters numbers.