An Algorithm Research And Implementation Of A Face Recognition System For Mobile

With the advent of the era of information intelligence,face recognition has gradually become one of the very valuable application directions in the field of computer vision,and therefore has received unanimous attention from both academic and industrial circles.Currently,the implementation of these face recognition technologies often depends on the hardware support of GPU,and the implementation of mobile face recognition still faces difficulties such as small storage space and insufficient computing power.In addition,face recognition systems for mobile devices are usually used in scenarios where users do not cooperate,so their performance is also adversely affected by problems such as facial occlusion.To address these problems,this paper investigates face recognition systems and designs and deploys a face recognition system for mobile.To address the key problem of poor face image quality in face recognition systems for mobile,an algorithm MTFLD,which combines face alignment and image quality evaluation into one,is proposed.the algorithm evaluates the current face image quality with an unsupervised method based on face landmarks detection,and selects high-quality face images for subsequent processes without increasing the system model size to improve face recognition performance.The experimental results also validate the excellent performance of MTFLD.Then,a robust face feature extraction algorithm OFRM is proposed for the masking problem in unconstrained face recognition,which adds a Mask Module that can dynamically generate a mask for the feature map based on the original face feature extraction algorithm,so that the network can focus more on the part of the face that is not occluded and ignore the features corrupted by masking,thus improve the performance of face recognition.The experimental results also fully validate the performance improvement brought by OFRM.Finally,this paper deploys the designed face recognition system on an Android device and performs a face recognition demonstration with good results.