Design And Implementation Of Personnel Managment And Control System On Mobile Devices Based On Partial Occlusion Face Recognition

As an important research branch in the field of computer vision,identity verification technology based on face detection and recognition has made considerable progress in recent years.This technology has achieved large-scale commercialization and popularization in many fields such as smart cities,traffic supervision and security monitoring,and has provided an important guarantee for the stability and development of the economy and society.However,in the non-restricted personnel control scene,partial occlusion of the face will cause the original structural features of the image to be lost,which greatly affects the accuracy of face recognition.In addition,with the widespread use of smart terminal equipment and the increase in computing power,the demand for personnel management and control based on mobile terminals has greatly increased.However,the current mainstream deep neural network model relies on the computing power of the server-side graphics processing unit.How to transplant the face recognition model to the mobile terminal to realize storage and real-time calculation is also a big challenge.The research focus of this paper is to design and implement a mobile terminal personnel control system suitable for partial face occlusion scenes based on lightweight face detection and recognition algorithms.First,we investigated and analyzed the current mainstream occlusion face detection and recognition solutions,and proposed a lightweight face detection model based on the MobileNet-V2 backbone network design based on the SSD target detection algorithm.The model integrates dilated convolution and feature pyramid network,which can greatly reduce model parameters and improve detection speed while ensuring the accuracy of face detection.In the research of facial feature extraction algorithm,this paper proposes a MobileNet-V2 network combined with a lightweight hybrid attention module for occlusion facial feature extraction.The network can reduce the information redundancy between channels,focus on the most informative part of the spatial feature map,make the model pay more attention to the feature learning of the non-occluded areas of the face,and have a smaller computational and parameter overhead.Secondly,we designed and implemented a personnel management and control system based on the above algorithm.The system integrates face detection and recognition models into a single mobile device,reduces image transmission costs and application response delays,decouples the calculation process from storage management and other processes,and can realize personnel management and control in an offline state.Finally,this paper tests the effect and performance of the proposed partial occlusion face detection and recognition algorithm through experiments,and verifies the overall function of the personnel control system.Experiments show that the algorithm proposed in this paper can achieve better occlusion face recognition results with less computational overhead.