摘要
由于指势可作为理想人机交互模式,研究指势识别具有重要意义,其中手指分割是关键.该文根据场景中任何可察觉的目标运动,都会体现在场景图像序列的变化中及彩色图像中红、绿、蓝三分量光强度在阴影区域存在差异,提出基于长序列多帧差分融合RGB彩色信息,建立自适应背景建模方法,从复杂背景中提取运动目标.根据手指在视频图像中的空间位置关系,提出自适应矩形结构元素对运动目标区域开运算,实现水平分割,以提取手指区域并确定手指尖位置.通过对不同背景的运动手指与指尖分割与提取,证实文中所提方法的鲁棒性.
Study of finger pointing gesture is significant due to the fact that pointing gesture is an optimal mode in human computer interaction. Extraction of fingers is a key problem to pointing gesture, Any variation of sensible motion target in scenes can be found and there are some differences between the intensities of R, G and B components in a color image in shadow region, An adaptive background modeling method is developed based on long series multi-frame difference and fusion of RGB color information. Motion targets are extracted from clutter background based on the obtained background. According to the relative spatial position of finger in the image, opening operation of structuring element with an adaptive rectangular shape is adopted to horizontally segment motion targets regions. Finger regions and position of fingertip are then determined. Experimental results show that motion finger and fingertip can be extracted efficiently from clutter background.
出处
《上海大学学报(自然科学版)》
CAS
CSCD
北大核心
2007年第4期421-425,共5页
Journal of Shanghai University:Natural Science Edition
基金
上海市教委与教育发展基金曙光项目(04CX72)
上海市青年发展基金资助项目(05AZ38)
关键词
分割
指势
背景
阴影
segmentation
pointing gesture
background
shadow