虚拟维修训练中的手势识别  

Hand recognition in virtual maintenance training

作　　者：汤志波[1] 张志利[1] 梁丰[1] 李向阳[1] 

机构地区：[1]第二炮兵工程大学导弹发射与定向瞄准技术军队重点实验室,陕西西安710025

出　　处：《计算机工程与设计》2014年第12期4278-4283,4288,共7页Computer Engineering and Design

摘　　要：为在虚拟维修训练中实现更加自然高效的人机交互,增强维修训练人员的沉浸感,提出一种基于Kinect深度图像和骨骼数据的手势识别方法。通过对维修样机和维修资源的维修特征进行研究,建立基于维修特征的维修动作集合,定义维修操作的常用手势库;在分析和构建手势识别系统流程的基础上,结合Kinect深度图像和骨骼数据实现手势图像的分割,基于形态学闭运算完成手势分割图的预处理,采用Zernike矩的方法实现对常用手势的识别。将该方法应用于机械设备的虚拟维修训练中,结果验证了该方法的可行性和有效性。To reinforce the training personnel’s immersion in the virtual maintenance training,a method for hand recognition based on the depth image and skeleton data from the Microsoft Kinect was presented to achieve more natural and efficient humancomputer interaction.Four sets of maintenance actions based on maintenance characteristics were established with the researches of maintenance characteristics of prototypes and resources in the virtual maintenance scene and the common gestures in the maintenance operation were obtained and defined according to these sets.A method for the hand segmentation combining the depth image with skeleton data was proposed on the basis of analyzing and designing the hand recognition system process.The pre-processing of hand segmentation image was completed using closed operation based on the morphology.The common gestures recognition was realized using Zernike moments.Finally the method was applied to the mechanical equipment’s virtual maintenance training and it is verified to be feasible and effective.

关 键 词：虚拟维修训练 手势识别 深度图像 闭运算 ZERNIKE矩 

分 类 号：TP391.9[自动化与计算机技术—计算机应用技术]