微型电脑应用2025，Vol.41Issue(7)：264-267,275,5.
基于肤色和可调式搜索窗口的小区门禁监控视频人脸识别
Facial Recognition for Community Access Control Monitoring Videos Based on Skin Color and Adjustable Search Window
姜正荣1
作者信息
- 1. 浙江迈新科技股份有限公司,浙江,宁波 315200
- 折叠
摘要
Abstract
A facial recognition method for community access control monitoring videos based on skin color and adjustable search window is proposed to improve facial recognition performance and achieve automated management of community access control.The method collects community access control monitoring videos and performs frame processing on them,uses a lighting com-pensation method to process each frame of video images to improve image quality.The processed frames of video images are converted to the YCbCr color space,the separation of brightness and color in the video images is completed.A Gaussian skin color model is used to model the color component and complete skin color region segmentation.Based on this,the size and posi-tion of the search window are dynamically adjusted to determine the facial candidate region images,and they are inputted into an improved convolutional neural network(CNN)based facial recognition model.Through deep learning and mining of facial fea-tures,more accurate facial recognition results can be obtained.The experimental results show that lighting compensation can improve the overall quality of video images.The proposed method can achieve facial recognition for community access control monitoring videos,the model loss value is about 0.10,AUC is 0.9245,and the time complexity is 15.61 s.关键词
搜索窗口/人脸识别/光照补偿/YCbCr颜色空间/高斯肤色模型/人脸候选区域Key words
search window/facial recognition/lighting compensation/YCbCr color space/Gaussian skin color model/facial candidate region分类
信息技术与安全科学引用本文复制引用
姜正荣..基于肤色和可调式搜索窗口的小区门禁监控视频人脸识别[J].微型电脑应用,2025,41(7):264-267,275,5.
