摘要
提出了一种以人的动作序列图像的轮廓为特征,基于RBF神经网络的日常行为识别方法。首先通过建立自适应背景模型分割出动作序列并提取轮廓,然后利用傅立叶描述子处理动作的轮廓线序列并进行数据维数压缩,结合径向基神经网络进行行为识别。实验表明,利用本方法对人的日常行为的正确识别率达到90%以上,具有简单、实用的特点。
A RBF neural network-based method for action recognition is presented by using the contours of image sequence as representative descriptors of human posture to achieve daily human actions recognition. The contours of an action sequences are segmented after an adaptive background model is built. Then the body contours are processed by Fourier descriptors with data dimension being reduced and the radial basis function neural network is employed to recognize human actions. The experimental results show that this method can achieve the correct recognition rate above 90%.
出处
《光电子.激光》
EI
CAS
CSCD
北大核心
2008年第12期1686-1689,共4页
Journal of Optoelectronics·Laser
基金
国家"863"计划:家庭服务机器人重点项目(2006AA040206)
留学回国人员科研启动基金资助项目([2007]24号)
关键词
行为识别
图像轮廓
傅立叶描述子
径向基神经网络
action recognition
image contour
Fourier descriptor
radial basis function neural network
作者简介
黄彬(1972-)男,博士研究生,研究方向为计算机视觉、模式识别等.E-mail:hb-sdu@126.com