Fishface算法研究及应用

针对人脸识别在学生刷卡系统中的应用要求,选用Fishface方法进行图像处理。处理过程:设有训练样本S,通过S找到一个可以降低识别复杂性和降低噪音的变换,把得到的标有名字的人脸图像经过这种变换后存储到数据库中。识别一个人脸图像时,将图像进行相同的变换处理,用得到的结果与数据库中的各个人脸样本比较,求出欧氏距离或者马氏距离,与最小距离相对应的那个人的名字就是输出结果。并将算法移植到嵌入式ARM系统中,实现学生人脸的动态采集和识别,对30个人脸样本测试后,识别率达到87.502%。

Research on the Fishface Algorithms and Its Applications

Aiming at the application requirements for the face recognition system of student credit card, the paper selects Fishface methods for image processing. That is ： setting a system for training sample S, by S to find a transformation to reduce the complexity of recognition and noise, then this face image to be marked with the name is stored in the database after conversion. When a face image is to be identified, the makes a same image transformation processing, compared with each individual face in database, find the Euclid- ean distance or Mahalanobis distance, then the person name that corresponding to the minimum distance is the result. The algorithm is also repotted to the embedded ARM system to achieve a dynamic collection of students and face recognition. The test with 30 individual face samples shows that the recognition rate is of 87. 502%.