三维手势跟踪是基于手势交互中的一个基础性研究课题,要实时、高精度地实现三维手势跟踪是一个具有挑战性的热点问题.为提高三维人手跟踪的精确度,提出一种基于多模型融合状态预测的粒子滤波跟踪算法.首先通过对基于数字手套的虚拟烤箱系统进行实验,结合人手的行为理解和描述建立了基于认知实验的人手运动状态预测模型；其次对人手跟踪过程中的数据建立人手运动模型,将Sigma点原理应用到人手模型数据上,得到基于局部分析预测模型；最后将这2个模型按照其与当前帧图像的相似度进行融合,得到粒子滤波过程中的状态预测模型.与退火算法相比较的实验结果表明,在运行时间基本相同的情况下,该算法通过改善粒子滤波过程中状态预测的精度提高了人手跟踪过程中的精度.%Three-dimensional hand tracking is important for gesture interaction to achieve high-precision 3D hand tracking in real-time. Targeting on high accuracy of hand tracking, a novel particle filtering algorithm by fusing multiple models of prediction is presented in this paper. First, with the analysis of the virtual oven system based on data glove, we construct a prediction model of human hand behavior based on cognitive psychology, by combining the behavior understanding and description of hand. Secondly, by constructing the hand motion model using the data in hand tracking process, and by applying Sigma point principle to the motion model data, we can get the local analyzing prediction model. Finally, we fuse the two models into a model according to the similarity with the current frame image, which is the state prediction model in the process of particle filter. Compared with the classic annealing algorithm, our algorithm improves the precision of 3D hand tracking.
展开▼
