基于三维图像动作识别算法研究  

Gesture Recognition Algorithm Research Based on Three-Dimensional Images

作　　者：李学相[1] 安学庆[2] 

机构地区：[1]郑州大学软件学院,河南郑州450002 [2]郑州大学数学系,河南郑州450002

出　　处：《计算机仿真》2013年第9期374-377,共4页Computer Simulation

摘　　要：人体运动时的行为特征具有多样性和复杂性,无法采用精准的形状基描述人体行为的不确定性,进而无法构建人体三维运动模型,不能对人体三维图像动作进行准确的识别。传统方法假设人体运动的形状基数量稳定,人体运动行为具有复杂的形变特征,导致形状基参数出现波动,采用固定的形状基无法准确分析人体的复杂运动特征,使得人体三维图像动作识别存在较大的误差。提出一种Murkowski距离三维图像动作识别算法,依据人体运动时的物理特征将人体运动结构形状基参数处于Murkowski距离约束条件下,采用非线性优化方法进行求解获取人体三维行为特征,获取人体行为三维模型,进而实现对人体行为的准确判断。实验结果表明,改进算法能够对人体三维图像动作进行准确的识别,并且具有较高的精准度。Due to the behavior charactefistcs of diversity and complexity of human movement,precise shape matrix is unable to describe the uncertainty of human behavior,thus unable to build three-dimensional human body motion model.The paper proposed a recognition algorithm based on three-dimensional image Murkowski distance movement.On the basis of physical characteristics of the human body movement in the structure of human body movement shape parameter in Murkowski distance constraints,the nonlinear optimization method was used to obtain three-dimensional human behavior characteristics and the 3-d models for human behavior,and realize the accurate judgment on human behavior.Experimental results show that the algorithm of 3-d image for human body movements has higher precision.

关 键 词：三维图像 人体行为 动作识别 

分 类 号：F127[经济管理—世界经济]