Research Of Real-time Human Face Detection And Recognition

In recent years, with the development of computer science, image processing, artificial intelligence, machine vision, pattern recognition and many other technologies, face recognition technology has attracted more and more attention and has been one of the most promising valuably areas of biometric technology, which is widely used in security, entertainment, education and many other areas. The related technology and research are in constant progress and development. The face recognition process consists of four stages including face detection, image preprocessing, feature extraction and identity recognition.Although the AdaBoost algorithm has a better performance in face detection, under the influence of the uncontrollable video environment and the user poses, it is likely to cause a lower detection rate, higher false alarm rate. Based on the AdaBoost algorithm, it uses the ASM algorithm to detect face region for further filtration. So it can do real-time face detection in video and get a better detection rate.In the image preprocessing stage, the paper uses image rotation, affine transformation and image cropping as the face image geometry normalization program, meanwhile, the illumination preprocessing program includes gamma correction, difference of gaussian filtering, equalization and hyperbolic tangent, which are used to solve the problems of uneven illumination and inconsistent size caused by the image collection and make the face image size keep consistent and weaken illumination effect.The Gabor wavelet transform has good advantages in biology background and visual characteristics, so it is used to extract facial feature. The facial feature is extracted by using Gabor wavelet, and then the FFT algorithm is used to accelerate feature extraction to solve computational efficiency caused by convolution operation and improve the system performance.In addition, the higher facial features dimension is one of problems caused by using Gabor to extract facial feature. The paper realizes face recognition based on Gabor+PCA+FLD method to take advantage of good dimension reduction capability of PCA and excellent linear discriminant ability to make up for their shortcomings. The experimental results show that the robustness of this face recognition algorithm is better and the recognition rate is ideal.Based on the proposed corresponding improvement programs and processes of the above stages, the paper has designed and implemented a real-time face detection and recognition system. The results show that the system has a better robustness for the human facial gesture, facial expression and illumination variation in images and achieves a higher real-time performance and gets a more satisfactory recognition rate.