A Face Detection Method Based On Gabor Filters And Improved BP Neural Network

The biological characteristic recognition is a kind of identification technology that uses the human's special physiology or behavior characteristic, it provided a kind of high reliability, good stability approach of identity appreciation. Face detection and Recognition is very popular branch of the biological characteristic recognition. And it is also a very active subject in the fields of the computer vision and the pattern recognition.Face detection and face recognition have a much widely and valued prospect of application in security recognition, personal identification and the work of inspection in public security organs. Face detection is the first stage of face recognition. The researchers always try to find a face location algorithm with high accuracy and efficiency for a long time. Presently, face detection has two kinds of method. One is based on the heuristic modeling method, and the other is based on statistical modeling method.In this thesis, a study on a large number of traditional methods to face detection is presented.In order to improve the rate of face detection, an new algorithm was presented based On Gabor filters and improved BP neural network. Improved Gabor filter is used to the training samples, and gets the face feature vector groups; In the detection step, firstly, the characteristics of intensity in eye region and symmetry in face were used for finding face candidates, uses the characteristics of spatial locality and orientation selectivity of Gabor filters to design eight orientation filters for extracting facial sample features from face images; Then, the feature vector based on Gabor filters is used as the input of the face/non-face classifier, and a reduced feature subspace is learned by Template Methods; Finally, the improved BP neural network is trained by the reduced features. The emulation experiments show that the methodg detecting rate is higher than traditional methods. Compared with the traditional face detection algorithm.The Advantage lies of this proposed algorithm in the following four areas.First, the overall image compared with the traditional Gabor.In this paper befor the Gabor filtering there is a feature location method,This method greatly reduces the time feature extraction.Second, compared with the traditional PCA dimension reduction,this paper involve the idea of drawing the template traversal.Using the principle of hierarchical block the feature vector dimension reduction.To overcome the traditional PCA method of vulnerability around the outside of the shortcomings.Third, compared with the traditional BP algorithm,This paper uses a variable step BP algorithm improvement.To complete the same number of iterations operation, improved BP network converges faster.Fourth, compared to tonal sthe traditiandard face database, This paper face database used in all handmade.Although lack of the professional standards of the face of the sample library, but with a more general meaning of film.