基于深度图的驾驶舱内飞行员动作识别

发布时间：2018-04-14 14:25

  本文选题：驾驶舱 + 飞行员　； 参考：《电光与控制》2017年12期

【摘要】：驾驶舱内飞行员工作量的计算需要飞行员的动作数量、动作时间等动作信息。目前,关于动作识别的研究一般都是对特定动作,如走、跳等的识别,无法应用于驾驶舱内的飞行员。同时,由于飞行员操作基本由手来完成,因此对飞行员的动作识别基本可以认为是对手部动作的识别。据此提出一种基于深度图的飞行员动作识别方法,该方法先通过对飞行员手部进行跟踪,再通过基于动作段的方法确定飞行员动作。此外还提出一种触发方法,以实现系统对动作的自动识别。实验结果显示,所提方法的动作识别率约为94.06%,表明该方法能够有效地识别飞行员动作。
[Abstract]:The calculation of the pilot's workload in the cockpit requires the action information such as the number and time of the pilot's action.At present, the research on motion recognition is generally about the recognition of specific movements, such as walking, jumping, etc., which can not be applied to pilots in the cockpit.At the same time, since the operation of the pilot is basically accomplished by the hand, the recognition of the action of the pilot can be regarded as the recognition of the hand action.Based on this, a depth map based action recognition method for pilots is proposed. The method firstly tracks the hands of the pilots, and then determines the actions of the pilots by the method based on the motion segments.In addition, a trigger method is proposed to realize automatic recognition of action.The experimental results show that the motion recognition rate of the proposed method is about 94.06, which indicates that the proposed method can effectively recognize the pilots' actions.
【作者单位】： 上海交通大学;
【基金】：国家自然科学基金(61305141)
【分类号】：TP391.41;V328
，

本文编号：1749721