Research On Face Detection And Recognition Method Based On Image

Face as one of the important characteristics of the human body, has a very strong distinction between constant and individual differences. It includes face detection, identification and tracing. Because of the clear advantages of human face, it has obvious advantages such as non-contact, whole process and moderate distance. It has been a research focus in the fields of pattern recognition, computer vision and artificial intelligence. Has a broad prospect of research in many and we have a stake in areas of national security, transportation, safe community, intelligent vehicles and other.In recent years, face detection and recognition has made a lot of achievements. Currently it has a lot of methods to quickly and accurately on the frontal face detection. However, in practice, in many cases because of poor image capture device to detect the position and is intentionally or unintentionally caused by uncooperative captured face image is not a positive image.In face recognition, due to the limited training samples, there will be low accuracy, slow running and poor robustness and so on. Therefore, there is still a lot of research value on human face image.In view of the above problems, this paper puts forward the corresponding improved algorithm in order to improve the rate of face detection and recognition rate. The important research contributions of this paper are described in the following points:(1) The detection algorithm of rotated face images. In this paper, an improved scale invariant feature transform(SIFT) rotated face detection algorithm is used. First, the principal component analysis(PCA) is combined with the SIFT method, and the rotation, translation, scaling and partial affine invariance of the PCA method are used to quickly complete the initial detection of the rotated face. Then, through the eye and mouth position, improve the detection accuracy. Finally, through the improved AdaBoost algorithm to train the classifier and calculate the key points matching rate, the accurate detection of rotated face. The experimental results show that compared with the traditional and new algorithms, the algorithm ensures the high detection rate and high efficiency.(2) Face recognition using improved HMM and RVM fusion method. First the original samples reduced dimension and feature extraction; then use HMM model by testing samples matching degree, to form a feature vector; finally using RVM to get feature vector for classification and recognition test, output the result of recognition. In the ORL face database verify the advantages and disadvantages of the algorithm, the experiment shows that the recognition rate of this algorithm has a good advantage compared to other algorithms, is a can be applied algorithm.(3) The face detection and recognition system in this paper is verified by a lot of experiments. Experimental results show that the system of light illumination, complex background, non frontal faces, different facial expressions change has very strong robustness and adaptability, in more complex environment can also keep high detection and recognition accuracy, and also had a very big increase detection and recognition rate.