Research On Face Detection Based On Skin Color Clustering And Segmentation

Human face is rich in unique biometric information,which is unique and highly identified,so it has been widely used as an authentication method in many important areas of life,such as intelligent surveillance,community security,financial payment and others.Face detection is the first step of face image information processing.It is of great significance and research value to accurately and rapidly detect and mark out the location and quantity of human face in a picture.The progress of science and technology in information security has been the challenge of the new technology,people gradually turned to the biometric authentication methods of digital higher security from the traditional methods,such as fingerprint,iris,face and other biological characteristics.Face has stability and uniqueness,high security,difficult to copy,as a biometric identification technology,has attracted more and more researchers' attention in recent years.In recent years,face detection has also made great progress with the rapid development of pattern recognition and artificial intelligence etc.At present,it has been realized the accurate and fast detection for the clear frontal face in normal environment.However,in real life,because of the influence of illumination,occlusion and complex background,the taken face image is not a clear frontal illumination,which will bring some problems to the subsequent processing of the face image.In view of current face detection algorithm some shortcomings,this paper based on the in-depth study of a large number of traditional face detection algorithm,and put forward relevant optimization algorithm,in order to improve the detection accuracy and robustness.The research contents are as follows:Firstly,in order to solve the problem of face detection at the complex environment of intense light irradiation and occlusion,this paper proposes a face detection method based on improved PSO(Particle Swarm Optimization)and K mean clustering skin color segmentation.First of all,this method put measured images transform into YCgCr color space,because the skin color information distribution more concentration in this color space,and then using improved particle swarm optimization and K clustering comprehensive method to skin segmentation.In order to remove the noise from the outside of the face area effectively and obtain the candidate face area,which also needs to be processed by the mathematical morphological and the face geometric features.Finally,the candidate face regions are verified by the improved AdaBoost algorithm.The experimental results show that the method has high accuracy and good robustness and adaptability.Secondly,aiming at the face image affected by uneven illumination,a new method proposed for multi faces detection based on skin color segmentation and feature location in this paper.The method detects whether exists color deviation of the image in the RGB space,firstly,and then uses the improved reference white algorithm for illumination compensation.The method detects whether exists color deviation of the image in the RGB space at first,if there has,then use an improved reference white algorithm for illumination compensation.Then,the preprocessed image is transformed into YCbCr color space for skin color segmentation,and the improved Ada Boost algorithm is used for face detection.Next,a priori knowledge obtained from a large number of human eye training experiments is used to mark the human eye.Finally,the human face image with eyes location is output.