Research On Key Technologies For Face Recognition

In recent years, face recognition technology is a hot topic in pattern recognition andartificial intelligence field, which has a very high value in theory and application of biometricidentification technology. Because it is easy to collect, uncontact and to conform to humanrecognition habit, etc, Authentication system based on face recognition is widely used inbody identification documents, access control systems and security systems.Noise is often existed in the process of getting the face image. Meanwhile, insufficientlight, uniform of the image size and gray are also existed in the process. All these features arenot conducive to face recognition. Firstly, denoising algorithm is carries on by using medianfiltering method which is effective to reduce noise interference. Then homomorphic filteringand histogram equalization on face image are used to weaken the light influence in facerecognition process. Finally, the face image is gotten which is uniform both in size and grayby rotating, shearing and scaling.According to the weight declaration in face detection of AdaBoost cascade algorithmwhich based on the features of Hear, the rule of weight update is improved by using thresholdmethod to control each weight changes. The improved algorithm has a high detect speed,high precision and a good real-time, and also is good for the facial feature extraction andrecognition.Because of the good frequency characteristic and direction selection characteristics,Gabor filters is widely used in the feature extraction of face images. But the data of extractionis too large. Therefore, Fisherface method is used to reduce the feature dimensions which areextracted from Gabor filters. For the impact of outliers point to strike the optimal projectiondirection for the training sample in Fisherface dimensionality reduction process, an adaptiveweighting method is used to solve the problem effectively. Face feature extraction based onGabor+Fisherface is not only fast, but also has a very high recognition rate. And it hasstrong robustness for the posture and facial expressions.Through the relevant theoretical research, a face recognition system is built on VC++platform. It is proved that the system has a high recognition rate, while its robustness is alsovery strong.