Flight action recognition is the base of objective assessment for flight training quality. The traditional flight action recognition method can't recognize the complex flight actions with randomness and fuzziness, a flight action recog-nition method based on Bayesian network is proposed. Firstly, the flight data curves of flight actions are clustered into some classes by hierarchical clustering based on DTW, according to their shape similarity. Then, descriptive features of fight data curves are selected to distinguish different shape according to the statistical analysis for dependence. Finally, the Bayesian network for flight action recognition is designed to fuse shape features and descriptive features. The flight action recognition experiment results show that the proposed flight action recognition method based on Bayesian network in this paper is effective with a high correct recognition rate.%飞行动作识别是客观评估飞行训练质量的基础.复杂机动动作具有较强的随机性和模糊性,传统的飞行动作识别方法难以有效识别.为此,提出一种基于贝叶斯网络的飞行动作识别方法.根据飞行动作中参数曲线形态特征,采用基于DTW距离的时间序列层次聚类方法进行分类;通过依赖统计分析方法确定参数曲线的描述特征;根据形态特征和描述特征构建贝叶斯网络;利用贝叶斯网络进行推理.仿真实验结果表明,基于贝叶斯网络的飞行动作识别方法对复杂机动动作具有较高的识别率.
展开▼
