视频监控下的行人性别检测  

Pedestrian Gender Detection under Video Surveillance

作　　者：苏宁 陈临强[1] SU Ning;CHEN Lin-qiang(College of Computer Science and Engineering,Hangzhou Dianzi University,Hangzhou 310018)

机构地区：[1]杭州电子科技大学计算机学院,杭州310018

出　　处：《现代计算机（中旬刊）》2018年第10期29-33,48,共6页Modern Computer

摘　　要：性别检测一直是计算机视觉研究和应用中的热门话题。目前性别的检测主要是设计有效的特征提取的方法提取人物脸部特征,然后利用分类器实现二分类。作为深度学习的重要组成部分,卷积神经网络在图像分类、语音识别等领域取得很好的效果。对于传统特征提取的方法在有些情况下并不能达到满意的效果,使用卷积神经网络与传统特征相结合的方法,提取行人头部与衣着信息进行训练进而检测行人性别。由对比实验结果可知,识别正确率可以达到85.5%,证明该方法的有效性。Gender detection remains an important task in the research and practical application of computer vision.Designing an efficient describing method to extract the face features and applying classifier to realize dichotomy is a popular way in the area of gender detection at present.Deep convolutional neural network has made great success on image and audio,which is the important component of deep learning.The method of traditional feature extraction can’t achieve satisfactory results in some cases.Using a combination of convolutional neural network and traditional features,it extracts pedestrian head and clothing information for training to detect pedestrian gender.The correct rate of recognition can reach 85.5%from the results of comparative experiments,which proves the effectiveness of the method.

关 键 词：特征提取 图像分类 卷积神经网络 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]