Research On Face Recognition Algorithm Based On Machine Learning And Design Of Attendance System

In recent years,China’s economy has developed rapidly,and many small and medium-sized enterprises have emerged.As the number of business units has increased,attendance has become an important part of the company’s personnel management.At present,biometric-based attendance systems are used in the market.,Attendance system based on face recognition is widely used due to its advantages of fast acquisition speed,high reliability and low cost.Face detection is the first and most critical step in face recognition.In order to solve the problem that Ada Boost algorithm needs to traverse the entire picture and cause low detection efficiency,this paper proposes an Ada Boost face detection algorithm based on skin color segmentation.This method first uses the clustering performance of the skin color of the face to detect and segment the picture,and then uses Ada Boost to detect the segmented area,which avoids the problem that traditional Ada Boost needs to traverse the entire image and cause low detection efficiency.Experiments show that the algorithm can improve the detection speed without affecting the accuracy of detection.For face recognition,the convolutional neural network in deep learning is applied to face recognition.The googlenet network architecture in the facenet deep network is used to fine-tune the network architecture,and the appropriate optimized loss function triplt loss is designed and selected.Finally,using tensorflow as a deep learning framework,face recognition was successfully implemented.The designed convolutional network was retrained to generate a network model,and each face was converted into a 128-dimensional feature vector.The network model was used for face pictures.sort.Finally,a set of face time and attendance punch-in system based on face recognition was built.This system uses VS2015 as the development platform,and uses opencv3.4 as the basic function library of the face information recognition module,to realize the administrator login,user registration,user management,attendance module,and report early warning and other functional modules.