摘要
针对OpenCV的HAAR特征分类器在人脸检测方面出现的识别效果和抗遮挡能力较差,误检、漏检以及动态识别实时性不强等问题,文中提出了采用Python+Dlib方法以及开源已训练好的大量人脸模型接口来实现人脸识别,对遮挡鲁棒性较好。通过对正脸以及偏转较小的图片和视频在人脸检测、特征点标定、特征向量提取以及人脸比对识别过程进行测试,实验结果表明,该方法在检测实时性、识别精度和效果上均优于OpenCV方法。
The HAAR algorithm of OpenCV has some problems in face detection,such as poor recognition effect,poor resistance to occlusion,false detection,missing detection and poor real⁃time performance in dynamic recognition images.In this paper,the method of Python+Dlib is adopted to implement face recognition with a large number of face model interfaces trained by open⁃source,it is robust to occlusion.Face and less deflected images and video were tested in face detection,feature point calibration,feature vector extraction and face comparison recognition process.The experimental results show that this method is better than OpenCV method in real⁃time detection,recognition accuracy and effect.
作者
张杜娟
丁莉
吴玉莲
ZHANG Dujuan;DING Li;WU Yulian(School of Health Services Management,Xi’an Medical University,Xi’an 710021,China)
出处
《电子设计工程》
2024年第1期191-195,共5页
Electronic Design Engineering
基金
陕西省教育厅专项科研计划项目(20JK0886)
西安医学院2021年校级科研项目(2021QN21)。
关键词
Dlib
人脸检测
特征点标定
特征向量
人脸识别
Dlib
face detection
feature point calibration
feature vectors
face recognition
作者简介
张杜娟(1981—),女,陕西西安人,硕士,讲师。研究方向:深度学习技术、图像技术等。
