基于改进逆向运动学的人体运动跟踪

Human motion tracking based on an improved inverse kinematics

随着人们对智能系统需求逐年增高,基于视觉的运动研究引起计算机视觉领域工作者更多的关注。这使它成为模式识别、行为学、行为处理分析与处理等学科的研究热门。现存算法存在需要标记、相机标定等各种约束条件,不能满足人们对人体运行跟踪的需求。论述了一种结合改进逆向运动学和图像模板匹配算法的人体运动位置的跟踪方法。该算法以改进逆向运动学为框架,首先依据逆向运动学知识与正向运动学知识计算出的关节点的粗略位置,对外观模型的各个模块进行模板匹配,接着确定关节点的最优位置,然后确定关节点的三维坐标值,最后重构得到三维动作序列。实验表明,在主观视觉感受与客观衡量标准两方面,此算法获得的实验结果都能够逼近乃至达到人体运动跟踪领域的最佳水准。

With the rising demand for intelligent systems,the study of vision-based human motion is drawing the machine vision investigators,making it become the research focus of pattern recognition,behavioral science,behavior analyzing and processing. The existing algorithms have many kinds of restriction conditions,such as marking and camera calibration,not being able to meet people＇s demand for tracking human motion. Therefore,this article proposes a human motion position tracking algorithm on the basis of video,combining the template matching and improved inverse kinematics. It first calculates the coarse position of joint point according to the inverse kinematics and forward kinematics,then applies template matching to each module of the appearing-model,and then determines the optimal location of joints and the 3D coordinates of joints,and finally obtains the 3D action sequences by reconstruction. Experimental results show that this algorithm can be close to and even reach the best level in both subjective visual feel and objective weighing standard in the field of human motion tracking.