摘要
为了解决快速、准确人脸识别的问题,提出了一种基于并行PCA算法的人脸识别方法。利用主成分分析法（PCA）能够降低特征维数、容易实现、训练时间较短的特点,设计实现了一种并行PCA算法,首先根据整幅图像提取出4幅部分人脸图像,然后将整幅图像和4幅部分图像同时由相同结构的PCA模型进行学习,提取人脸特征向量,通过欧氏距离进行测试图像与训练图像的匹配计算,最后通过测试图像与5级并行PCA模型的识别结果进行加权决策,从而实现人脸识别的目的。利用ORL人脸库的图像数据,在Matlab进行的仿真实验结果表明,该方法在准确性上有了很大程度的提升,识别的速度也相对较快,具有较高的鲁棒性。
In order to solve the problem of fast and accurate face recognition, a face recognition method based on parallel PCA al- gorithm is proposed. Using principal component analysis （PCA） method can reduce the dimension of features, easy to imple- ment, training time is short, the design and implementation of a parallel algorithm for PCA, first of all according to the whole im- age to extract the 4 part of face images, then the whole image and 4 partial images at the same time by the same structure of the PCA model of learning, face feature vector extraction, the Euclidean distance for matching calculation of the test images and training images, finally through the test image with the five level parallel PCA model identification results are weighted deci- sion, in order to achieve face recognition. Using the image data of the ORL face database, the simulation results in Matlab show that the method has a great degree of improvement in accuracy, the recognition speed is relatively fast, with a high degree of ro- bustness.
作者
赵亚鹏
ZHAO Ya-peng （College of Information Science and Engineering, Shandong University of Science and Technology, Qingdao 266590, China ）
出处
《电脑知识与技术》
2016年第7期147-148,155,共3页
Computer Knowledge and Technology
