Research On Face Recognition Technology In Embedded Environment

Face recognition is an important branch and research direction in the field of computer vision.It has been widely used in smart home,identity authentication and other fields.More and more academic members pay much attention to face recognition,which on the basis of machine learning.The main work of this paper is the research of face recognition technology,which focusing on the sub-pattern GLHist-LP face recognition algorithm,as well as its implementation,testing and analysis of the algorithm on the embedded platform with ARM.Based on the research and analysis of the existing face recognition technology,a sub-pattern GLHist-LP face recognition algorithm network structure is proposed.Compared with the existing face recognition algorithm,this structure reduces the dimension of face features,speeds up the implementation velosity of face recognition algorithm,and improves the accuracy and efficiency of face recognition.The main task of feature extraction is to form four different sub-patterns according to local face regions.The four sub-pattern images are convoluted by improved Gabor filter,and then the Gabor feature graph is obtained.The binary pattern histogram is used to reduce the dimension of the feature graph set and extract the 256-dimensional LGBP histogram information.The results are applied to the sub-pattern GLHist-LP face recognition algorithm,which reduces the main operation load of the auxiliary feature extraction algorithm.The auxiliary feature extraction is realized by utilizing LBP feature and its texture sensitivity and rotation invariance.In order to solve the problem that the main features lack the information of the overall facial structure,the dimension reduction operation is completed by using the 2D-PCA linear mapping algorithm to obtain the auxiliary feature with dimension 82.This operatioin can effectively increase the accuracy of the algorithm.This paper normalized the data of the main feature and the auxiliary feature,then the total dimension of feature increased to 338,thus the key feature information of face recognition is obtained.In addition,different weak Adaboost classifiers are trained to high precision and generalization recognition model by iteration and the algorithm is implemented on the embedded platform.This paper accomplish these tasks: main feature extraction,auxiliary feature extraction,feature fusion and classifier training achieve the function of face recognition.The sub-pattern GLHist-LP face recognition algorithm is based on the embedded platform.The test results show that the program runs smoothly,and the model works well.The accuracy of face recognition reach up to 94.46%,and the time of face recognition is less than 1 s,which achieves the desired goal and has practical application value.