中型组足球机器人传球动作辨识与再现  

Recognition and representation of ball-passing motions for a middle size league soccer robot

作　　者：田铠 卢振利[1,2,4] 徐惠钢[1] 顾启民[1] 毛丽民[1] 陈勇[1] 李斌[2] 

机构地区：[1]常熟理工学院电气与自动化工程学院,常熟215500 [2]中国科学院沈阳自动化研究所机器人学国家重点实验室,沈阳110014 [3]中国矿业大学信息与电气工程学院,徐州221116 [4]阿威罗大学电子和信息通信工程研究所,葡萄牙阿威罗3810-193

出　　处：《高技术通讯》2015年第6期614-621,共8页Chinese High Technology Letters

基　　金：国家科技支撑计划(2014BAK12B01);机器人学国家重点实验室开放课题(2014-008);校新引进教师科研启动项目(XZ1306);葡萄牙政府(FCT CIANCIA2007)资助项目

摘　　要：为了使中型组足球机器人具有识别并模仿人的传球动作的能力,在研究人的传球行为的基础上,提出了一种使中型组足球机器人能够识别和再现人的传球动作的可行方法。该方法利用Kinect感应器获取人体的三维骨骼信息,从生成的人体骨骼模型中选取有效表征人体传球动作的特征变量,然后通过阈值法和帧差法判定特征变量的变化,实现对人的传球行为的识别,并以此获取传球动作的传球角度,使机器人再现人的传球动作。针对人机交互中的传球动作辨识与再现提出了一种实际可行的解决方案。试验表明,该方法能够辨别多个角度传球的动作,并能保持较高的识别率,从而验证了该方法的实用性和有效性。In order to enable a meddle-sized league soccer robot to have the ability to recognize and imitate ball-passing motions, a realistic solutionfor recognition and representation of ball-passing motions of human players was put forward based on the study of the human behavior in ball-passing. The solution uses the kinect inductor to capture the three-dimensional human bone point information to create the human skeleton model, and selects some characteristic variables from the model to effctively represent human ball-passing. Then the threshold method and the frame difference method are used to determine the change of the characteristic variables to realize the recognition of human passing behavior, and based on this, the ball-passing angle is obtained to represent the ball-passing motion. The experiment proved that the proposed solution can accurately recognize the ball-passing motions from different passing angles. Thus the practicality and effectiveness of the method were verified.

关 键 词：中型组足球机器人 传球 动作辨识 动作再现 

分 类 号：TP242[自动化与计算机技术—检测技术与自动化装置]