摘要
单目深度图无标记人体姿态估计问题,由于动作的多样性,人体自遮挡,运动无规律等因素的影响,导致系统准确率低,鲁棒性不强和运行效率低。为此提出一种基于单目深度图点云的特征提取方法和回归方法,利用特征回归和关节点分类,可以在不使用时间信息的情况下,从单目深度图出估计出人体的关节点坐标。实验结果表明,与其他基于单目深度数据的姿态估计方法,以及相同情况下的多目方法比较,该方法的都能保持很好的精度。
Monocular camera mark-less pose estimation system suffers low accuracy, robustness and efficiency due to variety of action, self-occlusion of human body. A method of feature exaction from point clouds was proposed, in which a single-to-multiple(S2M) feature regressor and a joint position regressor were designed to quickly and accurately predict the 3D positions of body joints from a single depth image without any temporal information. Experiment result shows that the estimation accuracy is superior to that of state-of-the-arts and multi-camera based methods.
作者
陈莹
沈栎
Chen Ying;Shen Li(Key Laboratory of Advanced Control Light Process,Jiangnan University,Wuxi 214000,China)
出处
《系统仿真学报》
CAS
CSCD
北大核心
2020年第2期269-277,共9页
Journal of System Simulation
基金
国家自然科学基金(61573168)
关键词
计算机视觉
机器学习
像素分类
深度图像
人体姿态估计
点云
computer vision
machine learning
pixel classification
depth image
pose estimation
point clouds
