针对目前已有多目标威胁评估方法主观性强、稳定性弱、评估过程不连续的问题,综合考虑目标运动特性、目标行为意图,提出了一种基于动态贝叶斯网络(Dynamic Bayesian Network,DBN)和逼近理想解法(Technique for Order Preference by Similarity to an Ideal Solution,TOPSIS)的多目标威胁评估方法DBN-TOPSIS.通过分析目标特征指标间的节点关系,建立多目标威胁评估DBN.采用模糊理论,通过梯形隶属度函数对战场传感器、雷达等获取的连续型特征指标数据进行离散化处理,统一特征指标形态.利用联合树(Junction Tree,J-tree)算法进行动态威胁程度推理.构造DBN推理结果与TOPSIS评估矩阵之间的映射关系,采用TOPSIS法将威胁评估概率结果转换为威胁程度综合评估得分,进行多目标威胁程度准确排序.实验结果表明,DBN-TOPSIS多目标威胁评估方法具有较好的合理性和准确性.
A Multi-object Threat Assessment Algorithm Based on DBN and TOPSIS Methods
To solve the problems of strong subjectivity,weak stability,and discontinuous evaluation process in existing multi-object threat assessment methods,a multi-object threat assessment method DBN-TOPSIS based on Dynamic Bayesian Network(DBN)and Technique for Order Preference by Similarity to an Ideal Solution(TOPSIS)is proposed,taking into account the object motion characteristics and behavior intentions.By analyzing the node relationships between the object feature indicators,a multi-object threat assessment DBN is established.Using fuzzy theory,the continuous feature indicator data obtained from battlefield sensors,radars,etc.are discretized using trapezoidal membership functions to unify the form of feature indicators.The Joint Tree(J-tree)algorithm is used for dynamic threat level inference.The mapping relationship between the DBN inference results and the TOPSIS evaluation matrix is constructed,and the TOPSIS method is used to convert the probability results of threat assessment into a comprehensive evaluation score of threat level,and an accurate ranking of the multi-object threat level is conducted.The experimental results indicate that the DBN-TOPSIS multi-object threat assessment method has good rationality and accuracy.
万方数据
维普
