Research On Threat Assessment Method Of Battlefield Target Based On Bayesian Network

The battlefield is changing rapidly and the warplanes are fleeting.Threat assessment,as the basis for tactical decision-making,is a key part of "first enemy decision-making".The complex,uncertain and changeable battlefield environment makes it difficult for commanders to quickly and accurately assess target threats.Threat assessment based on artificial intelligence algorithms has become the current mainstream method.Compared with other artificial intelligence algorithms,Bayesian Networks have powerful uncertain information modeling capabilities and good interpretability.In order to accurately and efficiently assess the threat of battlefield targets,it is a prerequisite to quickly construct an accurate Bayesian network model.The main research contents of this paper are as follows:Firstly,for scenarios that require both speed and accuracy of threat assessment,with the purpose of ensuring the accuracy of the algorithm while reducing the time consumption,a dynamic programming structure learning algorithm of Bayesian Network,integrating the maximum weight spanning tree and improved MMPC,was proposed.It constrains the construction and traversal process of the dynamic programming parent graph by adding the dual constraints of the maximum number of parent nodes and the candidate parent sets,reducing the amount of scoring calculations and running time and storage of the algorithm.The experimental results showed that,compared with the traditional DP,SMDP,and MEDP algorithms,the algorithm in this paper has significantly reduced time consumption,scoring times,and storage capacity on the same standard network.Secondly,for special scenarios that have higher requirements for threat assessment speed,a dynamic programming BN structure learning algorithm based on node block order constraints was proposed,with the main purpose of improving network construction speed in further.The optimal parent sets of nodes obtained by traversing the parent graph constrained by the M-order matrix were applied to construct the initial graph structure.At the same time,the Tarjan algorithm,solving the strongly connected components of the graph structure,was used to obtain the node block order,which constrains the search process of the order graph in order to reduce the number of traversed node sequence paths and expand the scale of the network that the algorithm can handle.The experimental results showed that on four standard networks,the accuracy of this algorithm was reduced compared with SMDP,MEDP,BFSDP DPCMB and MMPCDP algorithms.However,the number of local scores,the number of traversed node sequence paths and the time consumption of the algorithm are significantly reduced.Thirdly,Bayesian Network was applied in the target threat assessment field,taking eight situational information,including target type,combat capability,and speed,as assessment indicators.The indicators were discretized to construct a BN-based target threat assessment model using the proposed algorithm and maximum likelihood estimation algorithm.Simulation experiments showed that for the same enemy target group,the built model can obtain a target threat ranking consistent with the relative entropy ranking method and the intuitionistic fuzzy inference method.When the target attribute information changes over time,a target threat assessment model based on dynamic Bayesian Network was constructed from defining state transition probability,in which the validity and rationality of the built model are verified by using the data of two time slices.Finally,the threat-assessment software based on Bayesian Networks was developed to display the image information from detection areas,current situation information,Bayesian Network threat assessment models,threat assessment results and other information.