Research On Face Recognition Based On Fisherface

With the innovation of technology,face recognition technology has gradually come to the real world.At present,there are many applications.Therefore,this paper analyzes and researches the face recognition technology,and completes the training and testing of a face recognition system based on Fisherface.We know that the face recognition process should include the production of face detection classifiers,face detection,face preprocessing,and training and recognition of face image sample sets.This article does the following work for the above process:First of all,several key improved LBP features are introduced.The feature extraction method based on MB-LBP and Haar-Like features is studied.According to the Haar-Like features of the face,combined with the AdaBoost algorithm,part of the AFLW face database is used as the sample set,and a personal face detector is trained by opencv,and the experimental comparison is carried out to get the best performance face detector.Secondly,due to the possibility of misjudgment when detecting human faces,the use of human face and human eye detectors in this paper works together to reduce the error rate of detection to a great extent and reduce the possibility of misjudgment.In addition,this article uses four consecutive preprocessing operations to reduce the effect of facial expressions,lighting conditions,and face in-plane rotation on recognition.Then,using ORL face database as a sample set and KNN as a classifier,the experimental results show that:the best discriminant feature spatial dimension has a great influence on the recognition rate of Fisherface algorithm,while the best description feature subspace dimension is only for PCA algorithm.The recognition rate has a great influence on the recognition rate of the Fisherface algorithm.The number of training samples has a significant impact on both algorithms.When there are few training samples,the recognition rate of the PCA algorithm is better than that of the Fisherface algorithm,but the overall look the highest recognition rate of Fisherface algorithm is 100%,which is better than the highest recognition rate of PCA algorithm by 96.25%.Finally,using OpenCV2.4.11 and Visual Studio 2013,a real-time face recognition system based on Fisher face algorithm is designed.This system has strong recognition ability for the targets collected in the training samples.Three individuals were tested here,and each person identified 22 consecutive times in a row and accurately identified the corresponding target number.However,because there are few training samples and there is no comprehensive illumination collection,the stranger face recognition is not good,and strangers are often erroneously identified as the target face in the training sample.Therefore,in order to obtain a good recognition rate,it is necessary to ensure that the facial image of the training sample set collects the omnidirectional lighting conditions,facial expressions,and the angle of the desired test.