摘要
研究人的步态识别优化问题,针对下肢关节点定位成为人体运动跟踪与分析中,存在精度不确切的问题。为解决上述问题,提出步态序列图像处理技术。首先从图像序列中提取人体目标,并建立下肢骨架模型和小腿轮廓模型;利用一种基于传统动态建模的圆周摆算法结合帧序列的连续性以及小腿轮廓数据自动检测膝关节位置和踝关节的粗略位置。通过仿真表明,小腿轮廓模型的匹配计算在踝关节预测点的周围搜索出精确位置,并验证了方法效果很好,为设计提供了依据。
In computer vision research field,the lower limb joints of human body location are one of the important research topics in the domain of human motion tracking and analysis,and many methods have achieved encouraging effects.First,moving silhouettes were detected by a background subtraction from image sequences,then skeleton model of lower limbs and silhouette model of crura were built.The knee joint position and approximate ankle joint position were obtained automatically by circle intersection pointing algorithm based on motion modeling with frame continuity and crura silhouette data.Finally,exact position of ankle joints was captured through matching computing based on silhouette model of crura.The paper lists experimental results,and shows that this method has satisfying performance.
出处
《计算机仿真》
CSCD
北大核心
2011年第2期284-287,324,共5页
Computer Simulation
关键词
下肢关节点定位
圆周摆
预测点
小腿轮廓
Lower limbs joint position detection
Circumference sways
Prediction point
Crura silhouette
