摘要
将傅里叶变换与边缘小波矩描述子相结合,提出一种人体行为动作的识别方法。凹凸复杂图像的质心到轮廓为非单一直线,据此,给出一种多段定向距离轮廓描述矩阵,实现轮廓特征的提取。分别对2类人体和4种行为动作进行仿真实验,结果表明,边缘小波矩描述子能较好地体现人体行为动作的形状轮廓局部特征,具有较高的识别率。
A method for respectively applying Fourier transform and edge wavelet moments descriptor to recognize human behavioral motion is proposed.In the process of contour feature extraction,contraposing not single line between centroid and contour of concave-convex complex images,a kind of directional distance contour description matrix with multi-lines is proposed.Some simulation experiments are done about two kinds of human bodies and four kinds of behavioral motion,experimental results show the edge wavelet moments descriptor has a good description and recognition rates to local feature of shape contour.
出处
《计算机工程》
CAS
CSCD
2012年第2期198-200,共3页
Computer Engineering
基金
江苏省博士后科研计划基金资助项目(1001027B)
江苏省高校自然科学研究基金资助项目(09KJB510002)
南京工业大学青年教师学术基金资助项目(39710006)
关键词
行为识别
小波矩
特征提取
轮廓描述矩阵
快速傅里叶变换
形状轮廓
behavior recognition
wavelet moments
feature extraction
contour description matrix
Fast Fourier Transform（FFT）
shape contour
