摘要
针对手势图像的肤色特点,结合肤色在RGB空间的阈值分割和YCbCr颜色空间上的聚簇特性,以及背景模型的应用,有效减少了背景中类肤色的干扰,完成了手部图像在复杂背景下的检测和分割;并采用图像的7个不变Hu矩描述子来表征不同二值化的手势轮廓;最后采用BP神经网络进行手势识别。实验结果表明该方法有较好的识别率和鲁棒性。
In light of the characteristics of skin colour in gesture images,the combination of threshold segmentation of skin colour in RGB space and cluster characteristics in YCbCr colour space as well as the application of background model effectively reduce the interference of similar skin colours in background and achieve the detection and segmentation of hand image in complex background.Seven constant Hu moment descriptors of image are used to characterise different binary hand gesture contours.At last,the BP neural network is applied to hand gesture recognition.Experimental results demonstrate that this method has higher recognition rate and better robustness.
出处
《计算机应用与软件》
CSCD
北大核心
2013年第3期247-249,267,共4页
Computer Applications and Software
关键词
背景模型
Hu矩描述子
BP神经网络
手势识别
Background model Hu moment descriptor BP neural network Hand gesture recognition
