Application Of Face Recognition Algorithm Based On Deep Learning On Home Service Robots

With the rapid development of artificial intelligence technology,home service robots are more and more widely used.Face recognition,as the main function of home service robots,has become one of the research hotspots in the field of computer vision and robotics.Traditional face recognition algorithms are limited to the laboratory environment.Face images in real environment and those taken by service robots will be affected by angle,light,resolution and shape,which will affect the accuracy of face recognition to a certain extent.In recent years,with the great progress of in-depth learning in image recognition,video recognition and other fields,it is of great significance to study the application of face recognition technology based on in-depth learning in home service robots.In this paper,the accuracy and speed of face recognition in home service robots are analyzed.The main work is as follows.(1)Summarize the advantages and disadvantages of traditional face recognition methods,and analyze the advantages of deep learning in face recognition,as well as the difficulties in the application of face recognition in home service robots.(2)In view of the shortcomings of open source datasets training model,such as face misrecognition,rejection recognition and poor generalization performance.It is concluded through research and experiments that excellent data sets have a great impact on model convergence and recognition rate.This paper collects and cleans millions of high-quality Asian face datasets,which lays a good foundation for model training.(3)In view of the low recognition rate of the original network model,this paper improves the Inception-ResNet-V1 network structure,increases the width and depth of the network,and obtains more prominent facial features to overcome the influence of angle,illumination and posture on the accuracy of face recognition.The experimental results show that the improved network structure training feature extraction model is used to validate on multiple testsets,and the recognition rate is greatly improved compared with the original network.(4)In view of the time-consuming feature extraction of Inception-ResNet-V1 network,which affects the real-time performance of face recognition system.This paper uses MobileNet lightweight network architecture as the backbone network of FaceNet.The experimental results show that compared with the original network,the modified network reduces the amount of computation greatly,and improves the speed of feature extraction effectively without losing recognition rate.(5)Face recognition system is built on Sun@Home Robot Platform of Beijing University of Information Science and Technology,and the algorithm is validated in real environment and related home service robot contests.The experimental results show that the face recognition system in this paper meets the requirements of recognition rate,real-time performance and stability.