Improved Face Tracking Algorithm Based on Block PCA and SVM

The recognition of portraits in surveillance video is different from static portrait recognition. Generally speaking, facial features in surveillance video cannot represent the original information. Therefore, face recognition is facing a great challenge, and there are certain requirements for the environment where information is collected. In order to continuously obtain the face in the video and find an effective face image, an improved face tracking algorithm based on the MEANSHIFT and block PCA+SVM is proposed. the Local Binary Pattern texture feature combined with color histogram is employed to improve the MeanShift algorithm. The experiments show that the accuracy rate of recognition has been greatly improved, from about 70% to about 90% compared with the traditional method. In addition, the improved method is more convenient than traditional method.