Research And Implementation Of Face Recognition System

With the rapid development of electronic technology and biological information technology,the pattern recognition technology based on biological characteristics has been widely used.Such as the Face Pay which provided by Ant Financial.The face recognition technology is one of rapid development and widely applied biometric identification technology,which is also a challenging and multi-disciplinary research subject.It contains the disciplines are: image processing and analysis,pattern recognition and machine learning,artificial neural networks and Mathematics and machine vision and artificial intelligence etc.Face recognition process mainly includes face detection,image preprocessing,feature extraction and pattern classification recognition.This paper mainly researches and discusses the theory and technology involved in face recognition based on BP neural network.The main research contents are as follows:The first step is face detection,image localization and preprocessing.Due to the face skin color stability and durability,it can be used for face detection and localization.In this thesis,the face detection algorithm use skin color as the first choice.The train of thought used YCbCr convert the face image to color space,then make skin color segmentation and remove existing noise,finally choose the face region.Image preprocessing adopts wavelet image denoising,histogram equalization to image enhancement and face image size normalization processing.Secondly,feature extraction of face images and training of BP neural network.The image is divided into edge region and non-edge region by edge detection with the features of faces,which obtain from the face detection testing experiment.And then according to the localization of the edge region,it can extract the facial features.Meanwhile it also used the PCA statistical analysis of facial features.The BP neural network was carried out to train the facial posture and facial features respectively.On this basis,this paper ascertains the parameter that affect the recognition rate,and select the optimized training parameters to improve the result.Finally,face recognition system is implemented using C#.With the data of pre-training,the recognition rate can be improved to more than 90%.And the expected test target is achieved.