Research On Face Recognition In Android

With the development of science and technology in China's economic development,artificial intelligence and pattern recognition have now become one of the main research directions of our society today.The mobile device based on Android also ushered in its highlight moment with the development of science and technology.Many applications already involve human personal information,personal privacy,and personal property.Therefore,the information security of mobile devices has become an urgent social problem to be solved.Face recognition as an effective means of identity recognition has become a relatively mature technology after rapid development in recent years.One of the core goals of face recognition technology is to solve the uncertainty of face images in the process of obtaining images using related devices,the diversity of face image patterns and the sample of face image data in the process of face image recognition Problems,such as the small sample of face data.This article mainly conducts related research on the above problems.In the process of face recognition,there are still many factors that affect the recognition effect in the operations of acquiring and processing face information.The direction and intensity of the sun's light will have a certain impact in the process of acquiring face images;the length of the beard left by the individual,the style of glasses worn,the hairstyle at different periods,and the artificial occlusion during the actual image acquisition process.Will have an impact.In actual application scenarios,the amount of data for training the face model is too small,and the face image is easily cracked by photos or videos to produce behavioral deception.It is precisely because there are factors that affect the accuracy of face image recognition and some factors that affect the safety of face recognition in the actual process,so face recognit ion technology still has a long way to go,and there are many important problems that we need to take one by one.To improve and solve.This paper will analyze the principles of face image recognition methods and the structural characteristics of related algorithms,and propose related solutions through a series of methods such as preprocessing of face images,convolutional neural network algorithms,and face image algorithms based on transfer learning.The preprocessing method can solve the uncertain problem in the face image acquisition.When we are collecting face images,a series of factors such as strong light,object occlusion,low light and shadow in real life will cause a lot of interference to the recognition of face images,so in the actual process we A pre-processing method is needed to adjust the face image.The adjusted face image maintains consistency in data such as image resolution,picture size,and other data,and to a certain extent,better image recognition results can be obtained.Convolutional neural networks and other methods have good results in dealing with facial expressions and facial signal noise.The convolutional neural network imitates a biological neural network of the human brain.It has the same self-learning ability of the network as some organisms,and imitates the perception characteristics of the biological network in some specific functions.And the convolutional neural network can effectively distinguish the boundaries of complex network classification,with multi-tasking capabilities,and its processed classification results quickly reach the ideal image classifier standard.Transfer learning is an important branch of deep learning methods.It can solve the problems that convolutional neural network model algorithms cannot solve to a certain extent.For example,small sample face data cannot make the neural network converge and the face data samples are insufficient problem.The main method is to solve the problem of different data samples but belong to the related recognition field through the existing knowledge network.In addition,compared with convolutional neural networks,transfer learning can save a lot of time and money,and can bring us great economic benefits.The main content of this paper is the different face recognition methods in the field of preprocessing of face images and the effect of face image recognition in the field of convolutional neural network algorithm models and transfer learning.Through improved face image recognition algorithms and some The mature face image algorithm performs related comparative experiments,and develops a face recognition system based on the Android platform,which includes functions such as registration and login of face images and unlocking of the Raspberry Pi on the hardware side.The experimental results show that training through migration learning The resulting model has fewer parameters and higher accuracy,which can improve the image recognition efficiency of the access control system.The system has low requirements on equipment and strong practicability.This article has carried out relevant experiments on the methods mentioned in the article,and has carried out relevant analysis and summary of the experimental results.According to the experimental results,the future development direction of face recognition technology is predicted and prospected.