【作者】 李小平； 赵德贵； 宋晔； 胡永亮； 刘琼昕； 付娜；

【Author】 LI Xiaoping;ZHAO Degui;SONG Ye;HU Yongliang;LIU Qiongxin;FU Na;School of Modern Distance Education,Beijing Institute of Technology;School of Computer Science and Technology,Beijing Institute of Technology;

【机构】 北京理工大学继续教育学院； 北京理工大学计算机学院；

【摘要】 针对细化人体轮廓得到的人体骨架图像颈部点处分支错移,导致人体骨架图像与实际人体不符的问题,提出一种基于骨架点分支数和线性方程连接的方法。该方法对骨架图像由左向右逐行地对每个骨架点计算其分支数目,然后根据分支的个数及位置情况确定人体头部、颈部及错移点坐标,消除头部与错移点间的骨架点,最后依据头部和颈部建立的线性方程计算确定需要被更新为骨架点的坐标,完成颈部修复。经理论分析和实验验证,所提出的方法能够有效地修复颈部分支错移,且骨架图像形状未变形,对人体骨架的分析具有良好的实用性。更多还原

【Abstract】 Because of wrong shift branch in neck,a skeleton image which is thinned by human body contour does not match with the actual human body.In this paper,it proposes a fixing method of neck wrong shift branch based on branches of skeleton point and connection of linear equation.The method scans the skeleton image up to down and calculates numbers of branches at each skeleton point.It determines coordinates of the human head,neck and wrong shift point by the number and location of branches.Then it eliminates skeleton points between the human head and wrong-shift point.Finally,it calculates coordinates of points that need to be updated to skeleton points by linear equation.The linear equation is established according to the head and neck point.Through theoretical analysis and experimental verification,the proposed method can effectively repair the neck wrong shift branch and the shape of human skeleton is not deformed,which provides an important foundation for the human skeleton key point extraction and behavior analysis.更多还原