Research On An Face Recognition Method Based On Dimension Reduction Of SIFT And BP Network

Face recognition technology is an important area research and hotspot in the biometric recognition. Ithas wide application prospect and high research value. Face recognition is that Making human facialfeatures as the label of the identity, face feature, which is include geometrical feature, color feature,statistical feature etc, is extracted using Artificial intelligence technology in order to distinguish differentface. However, human face recognition is often impacted by illumination, rotate, scale, or block, blur andother environmental factors on in the reality, although SIFT(Scale Invariant Feature Transform) algorithmcan solve the problem of face recognition in the changing environment, higher dimension makes itinefficient.To solve the problem in SIFT, this paper improves the SIFT, a kind of new method using BP networkis designed and it effectively improves the efficiency and accuracy rate of face recognition.Firstly, this paper analyzes research background and research status of the face recognition at homeand abroad, and studies the human face recognition technology,PCA algorithm and BP algorithm.Secondly, this paper proposes dimensionality reduction algorithm of the SIFT based on PCA, anddeeply studies the SIFT feature extraction algorithm. SIFT algorithm can solve environmental factorsinfluence on face recognition, because it has the characteristics of the scaling, translation and rotationinvariant. But due to the common SIFT algorithm with higher characteristic dimension, amounts ofredundant and irrelevant features, so its computation overhead is more. To solve the problem, the strongcorrelation characteristics, which are extracted by SIFT algorithm and, is picked up based on PCAalgorithm, new characteristics eliminate redundant features and reduce the dimension, so the recognitionefficiency of the algorithm is improved. The algorithm is validate in the Matlab platform. it effect that thealgorithm is effective.Finaly, This paper proposes the method of dimensionality face recognition based on SIFT and BPnetword: basic principle of BP network structure and BP algorithm is studied. The BP network structure ofthe SIFT dimension reduction features and implementing steps of BP algorithm are designed, and the BPnetwork as the classification recognizer and SIFT feature extraction based on PCA dimensionality reduction algorithm combining form a face recognition system, which is validated using the ORL facedatabase in the Matlab platform. the algorithm is validated in Matlab and proved effectively. Theexperiment proves that the proposed method in this paper improves the accuracy and efficiency of facerecognition.