机器人交互视觉中的多手势判断模型仿真  

Simulation of Multi Gesture Judgment Model in Robot Interactive Vision

作　　者：赵硕[1] 

机构地区：[1]齐齐哈尔大学,黑龙江齐齐哈尔161006

出　　处：《计算机仿真》2016年第7期410-413,共4页Computer Simulation

摘　　要：在对机器人交互视觉中的多手势判断建模优化的研究中,对一些多手势一起判断的过程需要快速的聚焦关键手势区域。由于采用当前的算法在建立机器人交互视觉中的多手势判断模型时,缺少快速聚焦过程,导致对多手势识别的训练的过程较为复杂,存在判断能力差的问题。提出改进光流场算法的机器人交互视觉中的多手势判断建模方法。上述方法先利用采集设备获取多手势图像,融合于YCbcr皮肤颜色模型,对手势区域进行分割,并利用光流场算法计算相邻帧图像间兴趣块的像素灰度差的总和,将计算的结果作为两帧图像间的光流场,利用主分量分析对获取的光流场进行降维去噪排序,在光流场序列中提取特征值较大的维构建多手势特征空间,以光流场列向量在多手势特征空间上的投影系数与手掌中心位置组成的混合观察向量为基础,组建机器人交互视觉中的多手势判断模型,仿真结果表明,改进光流场算法的机器人交互视觉中的多手势判断建模方法识别精确度高,鲁棒性强。In the paper, a modeling method was proposed to judge the multi gesture in robot interactive vision based on improved optical flow field algorithm . In the method, an acquisition device was used to obtain the multi gesture image firstly and integrated to YCbcr skin color model. Then segmentation of the gesture region was carried out, and the optical flow field algorithm was used to calculate the sum of the pixels＇ gray difference of the interested block between consecutive frame images. The results of the calculation were used as the optical flow field between two frame images. Principal component analysis was used to reduce dimension and rank the obtained optical flow field. In the optical flow field sequence, the dimension of the larger feature value was extracted to establish multi gesture fea- ture space. Based on the hybrid observation vector formed by the projection coefficient of the column vector of the op- tical flow field in multi gesture feature space and the center of the palm, multi gesture judgment model of robot interactive vision was formed, The simulation results show that the modeling method has higher recognition accuracy and stronger robustness.

关 键 词：多手势识别 光流场 手势分割 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]