Research Of Face Recognition In E-commerce Security Authentication

In recent years, e-commerce develops rapidly, in order to ensure the safety andeffectiveness in transaction, the security authentication problem has become a hotresearch field. As the first barrier of the network information security, identityauthentication is paid more and more attention. The method of traditional identityauthentication can only guarantee the identity reliability between computer andcomputer, it has little use to solve the natural isolation between human and computer.For resolve the problem, People pay attention to the technique of identityauthentication use the biological characteristics. Face recognition is considered as oneof the best technique which used for identity authentication for its unique advantage.On the basis of the general process to face recognition, this paper put it into theidentity authentication of e-commerce, then construct a model, and through theanalysis of this process, we get two key points: The first, the rapid face detection incomplicated background; The second, Face representation with the situation of singlesample.On the first, based on the study of classical AdaBoost, this paper make Pearsoncorrelation coefficient as a measure of the correlation between characteristics, andmake the AdaBoost with single threshold improved to dual threshold, then improvedthe traditional AdaBoos. Through the experimental results, it is confirmed reduce thealgorithm’s time and improve the detection accuracy. Based on the research of theface segmentation with double color space, combined with the improved AdaBoost,this paper proposed a new fitting method of face detection, and finally it is confirmedcan detect the face image in complicated background rapidly.On the second, this paper proposed a new picture rotation strategy. Through thesimulation of the four attitude of face, as turning left and right, looking up and lookingdown, combined with the translation, rotation, scaling and mirror transformation ofpicture, we construct a virtual samples set which has8virtual samples, to support thetraditional face representation algorithm, finally, the machining experiment validatesthe sample set.