摘要
目的提出一种基于图像分块和径向基函数(RBF)神经网络的人脸特征提取与识别方法,解决人脸识别中的高维、小样本问题.方法采用人脸图像的分块处理、奇异值分解压缩算法,降低特征维数,有效地解决了存储和传输中的数据压缩问题,运用基于聚类方法的RBF神经网络分类器进行人脸分类识别.结果通过实验和数据分析表明,该方法在人脸骨骼特征明显时具有较高的识别率,与基于整体人脸图像的识别效果相比,识别率提高了3%.结论笔者提出的识别方法具有良好的学习效率和识别精度品质指标.
In order to solve the problems of high-dimensional and small-sample-size in face recognition,the method of face feature extraction and recognition algorithm based on image subblock and radial basis function(RBF)neural network are applied in the paper.Human face images are handled by using the subblock processing and singular value and decomposition compression algorithm,so that characteristic dimensions are reduced and the problems of data compression in the course of storage and transmission are solved effectively.RBF neural network classifier based on clustering method is used for face images recognition.The method is of high recognition rate proved through experiments and data analysis while the human face being skeletal features evidently.Discrimination rate raised 3% compared with the recognition method based on the whole face image.The method of face classification and recognition based on image sub-block and RBF neural network is of a good learning efficiency and quality index of identification accuracy.
出处
《沈阳建筑大学学报(自然科学版)》
CAS
北大核心
2010年第3期586-591,共6页
Journal of Shenyang Jianzhu University:Natural Science
基金
住房和城乡建设部科研基金项目(2007-K03-04)
关键词
奇异值分解
RBF神经网络
人脸特征
分类
图像
singular value decomposition
Radial Basis Function Network
face recognition
classify
image