火力与指挥控制2025，Vol.50Issue(10)：101-110,10.DOI:10.3969/j.issn.1002-0640.2025.10.013
基于动态贝叶斯网络的代价敏感目标意图识别
Cost Sensitive Target Intention Recognition Based on Dynamic Bayesian Networks
摘要
Abstract
Accurate recognition of target intention is crucial for battlefield situational assessment.To address the issue of inconclusive reasoning due to varying sequence lengths in the process of intention decomposition and inference,this paper proposes a novel solution.This approach integrates dynamic Bayesian networks with a cost-sensitive mechanism and introduces a dynamic time warping algorithm tailored to optimize the assessment of intention confidence.In the initial stages where adversary intentions are not fully revealed,adaptive truncation of the decomposed intention sequences is employed.The similarity between these truncated sequences and the inference sequence is calculated,normalized,and used to enhance the precision of intention confidence calculations,thereby boosting the reliability of early-stage inference results on adversary intentions.Furthermore,the Three-way decision theory is introduced to manage the cost of misidentifying intentions,reducing the computational complexity of Three-way decision theory in multi-classification problems by standardizing cost calculations.Simulation results demonstrate that the proposed algorithm successfully combines dynamic features in intention recognition with cost sensitivity considerations.Additionally,it exhibits robust performance in scenarios with partially missing data,effectively mitigating the impact of incomplete data on recognition outcomes.关键词
意图识别/动态贝叶斯网络/代价敏感/3支决策/动态时间规整Key words
intent recognition/dynamic bayesian network/cost sensitive/three-way decision/dynamic time warping分类
武器工业引用本文复制引用
翟明圆,应淮冰,王武阳,何立栋,王晓玲..基于动态贝叶斯网络的代价敏感目标意图识别[J].火力与指挥控制,2025,50(10):101-110,10.基金项目
国家自然科学基金(61973163) （61973163）
江苏省自然科学基金优秀青年基金资助项目(BK20220104) （BK20220104）
