Face Recognition System Of Bank Service Robot

Face recognition belongs to the category of machine vision and also belongs to the category of biometrics,Face recognition is now the hot direction of artificial intelligence.The use of face recognition technology for identity verification has a stable and unique advantage,also by the face identification is very convenient and friendly and easy to be received.Face recognition technology has always been a challenging topic in the field of machine vision,and it is also a very widely used technology,especially in military security,public security and finance.In this paper,after studying the most advanced face recognition technology,this paper presents a real-time face recognition system based on different implementation methods and existing hardware conditions,and runs the system on the bank service robot.The essence of the process of face recognition is to do a series of characteristics of the process of face image,face recognition process of this subject mainly includes the following stages: detecting and locating the position of the face in the picture,obtaining the face picture,face picture comparison,through extensive research and comparison,the face detection algorithm with the highest face detection and the highest precision is selected,and also a better robustness and higher resolution convolutional neural network algorithm of face images comparison is selected.The main work of this paper is as follows:?The multi-view face detection algorithm based on funnel cascade structure is studied,and the principle of the algorithm is deeply studied and analyzed,and the training process and application condition of the face detector are studied,finally the algorithm is used in the face recognition system of the subject,and the fast detection speed and high detection rate are obtained.?The convolution neural network(CNN)is studied,the CNN model can be used to compare the two images,and the effect of depth on the performance of CNN model is studied,fnally it is found that the simpler and the deeper CNN model has better recognition performance.?On the basis of the existing hardware of bank service robot,the real-time face recognition system is implemented,and the C++ language is used to facilitate the transplantation of different platforms,then testing and evaluating performance of system and analyze,and finally the result of test are analyzed and compared.