摘要
智慧步道近年逐渐得到推广,多目标人脸检测是智慧步道的关键技术之一,这个问题目前还研究不足,特别是在角度偏移、部分遮挡、低分辨率情况下。本文针对智慧步道下多人脸识别场景,采用开源数据集,重点实验了基于FCHD神经网络的人脸检测模型,对该网络进行参数微调,然后利用虹软SDK和多线程技术开发了高效地人脸识别程序。实验表明:该模型对于多人脸场景具有较好的检测能力,即使对于角度有偏移、甚至有部分遮挡的人脸也能检测出来,系统具有高效的人脸识别效率。
Intelligent trails have been gradually popularized in recent years.Multi-face detection is one of the key technologies in Intelligent trails.This problem is still under-researched,especially in the case of angle offset,partial masking and low resolution.In this paper,based on the multi-face recognition scene in the Intelligent trails,the open source data set is used,and the face detection model based on FCHD neural network is tested.The parameters of the network are fine-tuned.The Arcface SDK and multi-threading technology were then used to develop an efficient face recognition program.Experiments show that the model has good detection ability for multi-face scenes,even if the angle is offset,even partially obscured faces can be detected.The system has efficient face recognition efficiency.
作者
唐晗
罗大晖
Tang Han;Luo Dahui(Changzhou Institute of Information Technology,Changzhou Jiangsu,213164)
出处
《电子测试》
2022年第15期63-65,133,共4页
Electronic Test
