摘要
利用Kinect深度传感器所获取的图像深度信息实现手部从背景中的分割,并通过在Hu矩基础上加入3个表达式,使不变矩包含更多细节特征,对手势目标进行识别,最后把得到的识别结果转换成控制指令通过Ad-Hoc网络传送给智能轮椅,实现运动控制.实验表明,通过该方法实现的手势以别存在图像旋转、缩放、平移的情况下具有不变性,对背景干扰具有较强的鲁棒性.
First,hands are split from the background based on depth information from images got by the Kinect depth sensor.And then,three formulas are added to the Hu moment to make invariant moment contain more detailed characteristics for hand gesture recognition.At last,recognition results are translated into a control command which is transmitted to the intelligent wheelchair through Ad-Hoc network,to achieve motion control.Experiment shows that the gesture recognition achieved through this method has invariance all in images＇ rotation,zooming,panning,and is of strong robustness to background interference.
出处
《机器人》
EI
CSCD
北大核心
2012年第1期110-113,119,共5页
Robot
基金
国家自然科学基金资助项目(51075420)
科技部国际科技合作计划资助项目(2010DFA12160)
重庆市科技攻关计划资助项目(2010AA2055)
关键词
KINECT
深度信息
HU矩
手势识别
智能轮椅
Kinect
depth information
Hu moment
hand gesture recognition
intelligent wheelchair
