Face Recognition Based On OpenCV And The Research Of Its Application In NAO Robot

In recent years,with the rapid development of the digital age,as a more secure technique in biometrics,face recognition has attracted more and more people's attention.Face recognition technology has many important applications in many fields such as smart access control,video surveillance,driver's driver's license verification,Alipay face,Mito Xiuxiu,housekeeping service robots,and virtual game players with real faces.This article is based on OpenCV(open source visual library)for face recognition research.Aiming at the Face recognition technology,this paper completed the experimental test of algorithm based on the exposition of theoretical algorithm,and accomplished the application of human-computer interactionthe in actual NAO robot.The main technologies include image preprocessing,face detection,face recognition and robotics applications.In the stage of image preprocessing,for the interference of various factors existing in the image acquisition process,this article introduces several common preprocessing methods,including image gray changes,image normalization,histogram equalization,and then these methods are all applied to specific experimental steps.In the stage of face detection,this paper use Adaboost algorithm to achieve.In the Adaboost algorithm,the face is represented by Haar-like features,the concept of the integral graph is introduced,and the face feature is calculated quickly by using the integral graph,which improves the efficiency of face detection.After that,Afterwards,this paper used the experimental samples in different situations to complete the precise positioning and detection tests.In the stage of face recognition and robot application,this paper firstly introduces the PCA-based face recognition algorithm and LBP-based face recognition algorithm.Secondly,the specific details of how to use OpenCV to implement the algorithm in the training and recognition process are given.Finally,the algorithm was tested and compared from the three aspects of the number of training samples,lighting changes,and rotation changes.It was concluded that the LBP algorithms has a relatively high overall efficiency,and then it was implanted on a humanoid NAO robot.Experiments show that the robot has a good application effect in face recognition,action interaction and voice greetings.