Svm-based Face Detection Algorithm Research

AS an important technology in dealing with the information of face , face detection has been an important research focus in the fields of computer vision and pattern recognition in recent years . The problems of Small-sample learning, Nonlinear classifier , Curse of Dimensionality and Local minimum point SVM can be solved better . So it will be very important to the research on face detection that if it is used together with staple face detection algorithm.This paper are concentrated on the following aspects :Then the face pass rate and non-face filter rate of the three different face detection classification filters : Template matching , PCA and liner SVM are analyzed firstly in experiment .The result is known that the performance of SVM which is trained by positive and negative samples is better than the performance of Template matching and PCA which is trained by positive samples. Then the strategy that face detection from coarse-to-fine is used .A SVM-based gray-scale image of the face detection algorithm is proposed. The algorithm is made up of three-tier classification structure . Average face template matching is used as the first layer classifier; linear SVM is used as the second layer classifier. The former two-tier classification are used as rough classifier, and the non-linear SVM is used as the core classifier . the face detection algorithm proposed is compared to the existing two-tier classification of filter face detection algorithm by the classifier with good training in experiment. Face detection rate and face detection speed are obviously improved, the number of error alarm significantly reduced when using Template matching plus linear SVM plus non-linear SVM three-tier classification face detection filter algorithm.The algorithm is proved effective and rebust.