The Research And Imrovement Of Face Recognition Algorithm

Present, although the face recognition technology, more and more people attention, anddomestic and foreign research has also made great progress, but in practical applications, due tothe lighting conditions, attitude, accessories shelter and face age changes in interference factors, aswell as the performance of the algorithm. So that the face recognition technology in practicalapplications there are still some drawbacks. To this end, this paper mainly the following aspects offace recognition algorithm based on global features and face recognition algorithm based on localfeature.Firstly, in the face image processing. In order to eliminate the impact of shootingconditions, face recognition algorithm, the geometry of the face image, gray normalizedpre-processing, the human eye positioning geometry normalized. This paper designs based on apriori knowledge of the human eye position coarse positioning, and then through the projectionpeak analysis of the human eye positioning algorithm, experimental results show that the fast eyelocation algorithm to ensure positioning accuracy on the basis of on improving the speed of thehuman eye positioning.Secondly, in face recognition algorithm based on the global characteristics of allsamples for PCA face recognition algorithm does not focus on the problem of the unified trainingleading to performance degradation and facial feature. Classification-based thinking, the sampleclassification training, to improve the performance of the algorithm can make the facial featurefacial feature subspace classification more focused, test results show that after the PCA transformfacial feature subspace, few eigenvalues larger feature vectors can represent the vast majority offace image feature, thus reducing the dimension of the feature subspace to improve theperformance of face recognition algorithms.Finally, in terms of face recognition algorithm based on local feature based on Gaborchange the face recognition algorithm based on local feature. Gabor transform in the frequencydomain and the time domain can be obtained best localized in the Gabor transform coefficientsdescribe the gradation characteristic of the human face image area near the given position, while having a light sensitive position advantages, the local characteristics of the human face that hasgreat advantages. The experimental results show that the cover in the face image by age change,the attitude disturbance factors, and still has a higher recognition rate.Based on global features face recognition algorithm based on local features facerecognition algorithms have their advantages and disadvantages and scope of this paper in thestudy and improvement of the human eye localization algorithm based on global features PCAface recognition algorithm and local features of face recognition algorithm based on Gabortransform, the test results show that the face recognition algorithm studied in this paper has adefinite improvement in recognition accuracy and performance.