基于遗传优化的自适应凸松弛人体姿势估计  

Human pose estimation: adaptive convex relaxation approach based on genetic optimization

作　　者：刘万军[1] 王伟[1] 付兴武[1] 曲海成[1] Liu Wanjun;Wang Wei;Fu Xingwu;Qu Haicheng(School of Software,Liaoning Technical University,Huludao Liaoning 125105,China)

机构地区：[1]辽宁工程技术大学软件学院,辽宁葫芦岛125105

出　　处：《计算机应用研究》2018年第1期313-316,共4页Application Research of Computers

基　　金：辽宁省教育厅科学技术研究一般项目(L2015216)

摘　　要：针对凸松弛方法在解决三维人体姿势估计的问题时存在迭代次数较多、准确度不高的不足,提出一种基于遗传优化的自适应凸松弛人体姿势估计算法。该算法首先对关键参数的更新方式进行自适应处理,然后利用遗传优化算法对该关键参数的初始值进行寻优,最后利用寻优结果对凸松弛方法中闭式解的公式进行调整。实验结果表明,提出的算法迭代次数更少,准确度更高,更有利于实际应用。In order to overcome the drawbacks of large number of iterations and low accuracy of convex relaxation approach during solving 3D human pose estimation,this paper proposed an adaptive convex relaxation approach based on genetic optimization.Firstly,it adaptively updated the key parameter.Then it used the genetic algorithm to optimize initial value of the key parameter.Finally it improved the expression of closed-form solution in the convex relaxation approach by using the result of optimization.The experimental results show that the improved algorithm has less iterative numbers but higher accuracy,which can be much helpful to practical applications.

关 键 词：姿势估计 凸松弛 自适应 遗传优化 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]