A Study Of Applications Of SIFT In Face Recognition From Video Surveillance

Video surveillance has been widely used in daily life so that face recognition based on video surveillance has become a hot topic at home and abroad. And it also is a must in such an international security situation. However, the real video surveillance environment is low-resolution, multi-angle, uncontrolled environmental condition and etc. which are challenges for face recognition. SIFT as a advanced algorithm with excellent robust and separability has features like scale invariance, rotation invariance and not sensitive to light and shade that is what needed by video surveillance face recognition.Tues, this research explores the application of face recognition using SIFT in video surveillance, including applying SIFT in video surveillance face recognition with single training sample and face re-recognizing. Evaluate on SCface database and compare with the traditional PCA method, SIFT does improve the recognition performance. And in the face re-recognizing application, the correct recognition rate can reach 95%.Direct at the wrong match situation when using different parameters, this paper gives a algorithm of SIFT in video surveillance face recognition based on eye blocking. Combine with pre-processing, the algorithm divide the image into two blocks according to the eyes line after feature extracting. Finally, match key points in corresponding blocks dividedly for reducing the wrong match. Evaluated on SCface database, it showed that the proposed approach could avoid the wrong matches between top block and bottom block which enhances the accuracy of the algorithm.Besides, in order to promote recognition performance, here also comes up with a algorithm of SIFT in video surveillance face recognition based on 3D by single sample. In the proposed algorithm, firstly a 2D frontal face with high-resolution was taken to build a 3D face model. And then several virtual faces with different poses were produced from the 3D face model. At last, both the original frontal face image and virtual face images were put into gallery set. Except evaluating on SCface database, this paper also build a LABface database for experiment. The result showed that the proposed algorithm could effectively improve the recognition performance in video surveillance face recognition using SIFT.