视频中运动人体骨架提取方法  

Algorithm of Moving Human Skeleton Extraction in Videos

作　　者：吴翊萱 贺俊吉 WU Yi-xuan, HE Jun-ji （Logistics Engineering College,Shanghai Maritime University,Shanghai 201306, China）

机构地区：上海海事大学物流工程学院,上海201306

出　　处：《电脑知识与技术》2017年第4期151-153,共3页Computer Knowledge and Technology

摘　　要：提出一种基于ZHANG和SUEN快速并行细化算法（即ZS细化）的人体骨架提取方法。首先使用高斯混合建模背景差分法提取运动人体二值前景图像,再利用ZS细化算法对人体二值图像进行初步细化,最后使用8个模板对ZS细化结果进行单像素化得到单像素宽的人体骨架模型。实验结果表明,该方法能成功提取单像素宽的人体骨架,简单高效,针对单幅图像提取骨架平均耗时仅需0.3秒。An method to extract human skeleton based on ZHANGSUEN parallel thinning algorithm（i. e., ZS thinning） was proposed. Firstly, background subtraction by Gaussian mixture modeling was applied to get the moving human binary image, then ZSthinning was applied to thin the human binary image preliminarily.Eight templates were used to do further thinning to get the human skeleton model of single-pixel width. Experimental results show that the proposed method can extract the human skeletonmodel effectively. The algorithm is effective and the average consuming time to extract the skeleton of an image was only 0.3s.

关 键 词：高斯混合建模 ZS细化 单像素宽 骨架模型 

分 类 号：TP391[自动化与计算机技术—计算机应用技术;自动化与计算机技术—计算机科学与技术]