基于改进Faster R-CNN算法的行人检测  

Pedestrian Detection Based on Improved Faster R-CNN Algorithm

作　　者：单志勇[1] 张鐘月 Shan Zhiyong;Zhang Zhongyue(School of Information Science and Technology,Donghua University,Shanghai 201600)

机构地区：[1]东华大学信息科学与技术学院,上海201600

出　　处：《现代计算机》2021年第23期124-128,共5页Modern Computer

摘　　要：随着科学技术的发展和社会智能化技术的进步,行人检测技术成为各类监控场所必要的监控技术。利用基于Faster R-CNN的行人检测算法,使用DetNet特征提取网络从输入的图片中提取出检测目标的特征,区域建议网络RPN生成建议候选区域,利用检测网络对行人目标进行识别和定位。采用ROI Align解决ROI Pooling两次量化问题,结果表明:该行人检测算法在USC行人数据集和INRIA数据集等公开行人检测数据集以及道路采拍的数据集上检测有很好的效果。With the development of science and technology and the progress of social intelligent technology,pedestrian detection tech⁃nology has become a necessary monitoring technology for all kinds of monitoring places.Using the pedestrian detection algorithm based on Faster R-CNN,the DetNet feature extraction network is used in the experiment to extract the features of the detection target from the input image,the region recommendation network RPN is used to generate the recommendation candidate region,and the detection network is used to identify and locate the pedestrian target.In the experiment,ROI align is used to solve the problem of twice quantify⁃ing ROI pooling.The experimental results show that the pedestrian detection algorithm improves the detection efficiency and accuracy on USC pedestrian data set and INRIA data set.

关 键 词：Faster R-CNN 行人检测 DetNet ROI-Align 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]