摘要
为真实再现增强现实环境下全息沙盘异地协同标绘的交互过程,提出了一种基于视频的异地协同标绘动作重构技术.使用单目RGB相机采集作业人员的实时视频信息,基于YOLO实现多人检测,并确定作业人员在三维空间中的相对位置关系;基于改进的ResNet网络对检测到的每个作业人员进行关节点的三维位置预测;然后将估算出的本地作业人员的关节位置实时传送到异地,经过逆向运动学及低通滤波器的优化驱动虚拟化身的运动.实验结果表明,能够准确地再现异地作业人员的标绘动作,满足实际应用需求.
In order to realistically reproduce the interactive process of collaborative plotting of holographic electronic sand table under the augmented reality environment,a video-based distributed collaborative plotting reconstruction method was proposed.Firstly,the operators were tracked based on YOLO using a single RGB camera simultaneously to estimate the relative position of those people in 3 D space.And an improved ResNet network was used to predict 3 D joint positions for every operator.Then the estimated joint positions of the local operators were sent to other place to drive the avatar’s movement in real time through low-pass filter and the method of inverse kinematics.The experiment results show that the proposed method can reproduce the operator’s motion and meet the application requirements.
作者
李维浩
姚世明
李蔚清
苏智勇
LI Wei-hao;YAO Shi-ming;LI Wei-qing;SU Zhi-yong(School of Automation,Nanjing University of Science and Technology,Nanjing,Jiangsu210094,China; School of Computer Science and Engineering,Nanjing University of Science and Technology,Nanjing,Jiangsu210094,China)
出处
《北京理工大学学报》
EI
CAS
CSCD
北大核心
2019年第12期1298-1303,1320,共7页
Transactions of Beijing Institute of Technology
基金
“十三五”装备预研资助项目(61409230104,315100104,41401010203)
中央高校基本科研业务费专项资金资助项目(30918012203)
上海航天科技创新基金资助项目(SAST2019009)
关键词
协同标绘
多人检测
ResNet
三维姿态估计
动作重构
collaborative plotting
multi-person detection
ResNet
3D pose estimation
motion reconstruction
