摘要
采用Part Affinity Field(PAF部分关联域)与卷积神经网络(CNN)结合的模型,解决深度图像下人体姿态估计问题。首先,通过CNN得到人体的一组特征图。然后,使用CNN分别提取其关节点信息以及PAF信息。最后,采用图论的匹配方法对各个关节点进行推理,将同一个人的关节点连接起来得到估计结果。实验结果表明,文中方法可以很好应用于深度图场景下。
A model is proposed to solve the problems of depth map multi-person pose estimation.The model is composed of Part Affinity Field and Convolutional Neural Network.Firstly,a set of human body features is obtained by using Convolutional Neural Network.Secondly,Convolutional Neural Network is used to extract its keypoint information and PAF information respectively.Finally,the graph theory matching method is used to infer the key points,the key points of the same person are connected to get the result.The experimental results show that the proposed method can be applied to depth map scenes.
作者
刘涛
杨璐
邵肖伟
LIU Tao;YANG Lu;SHAO Xiaowei(Tianjin Key Laboratory for Advanced Mechatronic System Design and Intelligent Control(Tianjin University of Technology),Tianjin 300384,China;National Demonstration Center for Experimental Mechanical and Electrical Engineering Education(Tianjin University of Technology),Tianjin 300384,China;Center for Spatial Information Science,University of Tokyo,Kashiwa 2778568,Japan)
出处
《智能计算机与应用》
2020年第1期103-108,共6页
Intelligent Computer and Applications
基金
天津市自然科学基金(16JCQNJC04100).
关键词
部分关联域
卷积神经网络
深度图
人体姿态估计
图论匹配
Part Affinity Field
Convolutional Neural Network
depth map
person pose estimation
graph theory matching
