基于代价敏感多类别三支决策的目标意图识别方法

A target intention recognition method based on cost-sensitive and multi-class three-branch decision

针对传统意图识别算法在现代化空中对抗中识别飞机行动意图结果存在冲突,甚至无法识别目标意图的问题,提出了一种基于代价敏感多类别三支决策的目标意图识别方法;建立意图识别与代价敏感多类别三支决策相结合的数学模型,在每一个识别阶段计算出误分类代价损失值最小的那一种意图分类,得到唯一的识别结果,避免冲突结果的产生,同时,引入误分类代价第二最小损失的分类结果作为延迟域的备选选项,弥补多类别三支决策延迟域缺失的这一状况。最后通过仿真实验,验证了设计算法的有效性。

Aiming at the problem that the traditional intention recognition algorithm has conflict in recognizing the intention of aircraft operations or even fails to recognize the intention in modern air combat, this paper proposes a target intention recognition method based on cost-sensitive and multi-class three-branch decision. A mathematical model combining intention recognition and the cost-sensitive and multi-class three-branch decision is established. At each recognition stage, intention classification with the lowest cost loss value of misclassification is calculated, and the unique recognition result is obtained, which avoids the conflict result. At the same time, the classification result with the second lowest cost loss of misclassification is introduced as the alternative of delay domain, which makes up for the absence of delay domain of the multi-class three-branch decision. Finally, the effectiveness of the design algorithm is verified by simulation experiments.