Face Recognition Algorithm And Its Implementation On The QT Platform

Nowadays, face recognition is involved in the fields of biology, machine learning and so on. It is important to combine the physiological features of human face with the computer. Many face recognition algorithms often depend on the computing power of the operating platform, and the real-time performance is poor. Taking the efficiency of face recognition as the main line, this paper makes some optimization and improvement on some steps in the process of face recognition.In the process of face recognition in facial feature extraction is an important part of face recognition, efficient face recognition classification the key lies in whether or not to extract accurate representation of face features, but in the feature extraction process in the sample picture dimension are generally very high, difficult to efficient extraction of face image feature, such as image matrix 112* 92, then covariance matrix is 304 10* 10 304, which will causes the program to run for a long time. In this paper in order to solve this problem using a matrix of image segmentation method, is both ears, face and so on. After the segmentation, the 6 matrix is obtained by the improved PCA algorithm for feature extraction, and finally the SVM features of the classification and recognition.Proved by the experiment using this algorithm only face samples of 65%of critical data can also get higher recognition accuracy, and because the target matrix is relatively small so the recognition process speed also than the traditional PCA+SVM method faster.Finally, a face recognition algorithm based on PCA and SVM fusion algorithm is proposed. Portable face recognition system based on Linux and Samsung Exynos4412 hardware platform, and using OpenMP multi-threaded optimization, to achieve millisecond recognition speed.