Research And Implementation Of Face Recognition Method For Guiding Robot

With the rapid growth of security inspection and financial and trade applications,biometric technology has received new attention,and face recognition is one of the most widely used technologies in all biometric methods.Face recognition has gradually become a research hotspot in the field of artificial intelligence because of its great application prospects in financial verification,medicine,human-computer interaction and public security.In this paper,the greeting robot platform as the entrance,combined with computer vision technology and motion control technology,not only enables the robot to have the ability of face feature recognition,but also combined with tracking algorithm,realizes the face tracking servo system,and enhances the human-computer interaction ability of the robot.The main work of this paper is as follows:Firstly,the basic structure and training method of convolutional neural network based on deep learning are introduced,and its core working principle and main optimization schemes are introduced.Nowadays,in order to improve the network performance,the mainstream methods are expanding the network structure or increasing the amount of training data,which makes the amount of training calculation increasing.In this paper,a new training method based on mixup is used to optimize the training mode of the network,which reduces the error rate of image classification by about one percentage point and improves the performance of the network.Secondly,the process of face feature recognition is introduced,which mainly includes three main steps: face detection,face correction and feature recognition.The advantages and disadvantages of traditional detection methods and convolutional neural network-based methods are discussed,and the most suitable scheme is found through actual operation verification.At the same time,the network training method based on mixup is applied to the network training of face feature recognition,which proves that the accuracy of face recognition is indeed improved.In order to increase the level of human-computer interaction,four recognition functions,facial expression,head posture,gender and age,were added to the training.Finally,in order to enhance the level of intelligent interaction of the robot,we add the tracking function of the robot's target face.At the same time,we apply the trained face feature recognition network to the greeting robot to form a complete system.Different recognition tasks can be accomplished by voice control robot,and the actual operation also achieves better accuracy and real-time.