摘要
为了解决当前实际场景人脸图像特征难识别,人脸图像库数据庞大而导致标记工作压力大的问题,本研究基于自动标记解决方案,对大数据人脸样本库进行自动标记,基于深度神经网络框架,对自动标记结果进行学习训练,形成精准识别机制。首先,采集大数据人脸图像,为人脸识别做好数据训练准备。然后,结合直方图均衡化和OpenCV训练特征函数,实现人脸检测,达到提高标记工作效率和质量的目的。最后,从分析深度神经网络原理出发,有机组合隐藏层的激活函数,基于ReLU激活函数和优化训练策略,建立深度神经网络人脸识别机制。实验测试结果显示:本文算法有利于人脸高精度识别系统的落地,为智能人脸识别设备奠定算法基础。
In order to solve the current problems of difficulty in recognizing face image features in actual scene,and the great pressure on marking resulting from the huge face image database,this study conducts automatic marking of big data facial image sample library on the basis of automatic marking solution,and drills the learning of automatic marking results on the basis of deep neural network framework in order to find out a precise recognition mechanism.First of all,it collects big data facial image to prepare for the drill of face recognition data.Then,by integrating histogram equalization and opencv training feature function,face detection is realized in order to improve the efficiency and quality of marking.Eventually,by analyzing the principle of the deep neural network,and integrating the activation function of hidden layer,the deep neural network face recognition mechanism is established on the basis of Relu activation function and the optimal training strategy.Experimental results show that the algorithm in this paper is conducive to the landing of high-precision face recognition system,and lays the algorithm foundation for intelligent face recognition equipment.
作者
赵卫东
秦锋
ZHAO Weidong;QIN Feng(Chuzhou Vocational and Technical College,Chuzhou 239000,Anhui;Anhui University of Technology,Maanshan 243002,Anhui)
出处
《攀枝花学院学报》
2020年第5期67-73,共7页
Journal of Panzhihua University
基金
滁州职业技术学院重点应用研究项目(YJZ-2019-43)
安徽省职业与成人教育学会2018年教育科研规划课题(AGZ18071)。
