论文部分内容阅读
在虚拟现实环境中手部跟踪是视觉交互系统的基础和核心。针对现有视觉跟踪方法在手部运动姿态、尺度变化及复杂背景条件下出现的稳健性等问题,结合纹理和轮廓信息,利用基于梯度方向局部二值模式特征为基础的粒子滤波跟踪算法,建立局部和全局的特征直方图描述,实现手部跟踪。针对粒子匮乏问题,利用红外深度信息,并引入基于群智能的人工蜂群算法,将当前时刻的观测信息融合在粒子预测的采样和更新阶段,高效完成目标的搜索和优化,降低粒子集衰减程度,改善状态估计精度。实验结果表明,该方法在各种复杂背景下可以实现手部的稳健跟踪。
Hand tracking in the virtual reality environment is the foundation and core of the visual interaction system. Aiming at the problem that the current visual tracking method appears in the hand gesture, the scale change and the complex background condition, combined with the texture and contour information, the particle filter tracking algorithm based on the local directional binary mode feature is established. Local and global feature histograms are described for hand tracking. Aiming at the problem of particle shortage, the infrared depth information is used and the artificial bee colony algorithm based on swarm intelligence is introduced to integrate the observation information of the current time into the sampling and updating stage of particle prediction, to efficiently accomplish the target search and optimization and reduce the degree of particle subset attenuation , To improve the state estimation accuracy. The experimental results show that this method can keep the hand robustly tracked under various complicated background.