摘要
针对人脸在整个图像中识别的问题,提出了基于神经网络的前端人脸检测系统。该系统首先将训练集中每个图像划分成独立的小窗口,并标示出每个窗口中包含的人脸图像,并据此对神经网络进行训练。在利用神经网络对人脸进行识别的过程中,对待检测图像划分为分辨率为金字塔形的小图像,然后对每个分辨率的小图像进行人脸识别。同时为了提升系统对人脸检测的准确率和误报率,需要对多个神经网络输出的阈值进行权衡。本系统使用自适应算法,增加了错误检测集以提升系统的检测精度。
To handle the problem of detecting face from an entire image,a neural network-based face detection system is present. Firstly,the image in the training set of the neural network is divided into small images,and the ones which contain the human face are marked out. Then use the training set to train the neural network. During the period of face detection by using the neural networks,the images which is to be detected will be separated into different resolution,which like a resolution pyramid,and detect human face in the small images. Meanwhile,in order to improve the accuracy of the face detection and lower the false alarm rate,the output of multiple neural networks should be balanced. The system uses the adaptive algorithm to increase the set of error detection and then enhance the detection accuracy of the system.
出处
《激光杂志》
北大核心
2015年第3期54-57,共4页
Laser Journal
基金
重庆市科委自然科学基金计划(2010BB2244)
重庆市教委科研项目(KJ131302)
关键词
人脸识别
模式识别
计算机视觉
人工神经网络
机器学习
Face detection
Pattern recognition
Computer vision
artificial neural network
Machine learning