基于改进Faster R-CNN的目标检测算法研究  

Research on the target detection algorithm based on the improved faster R-CNN

作　　者：张海明 史涛 ZHANG Haiming;SHI Tao(College of Electrical Engineering,North China University of Science and Technology,Tangshan 063210,China;College of Electrical Engineering and Automation,Tianjin University of Science and Technology,Tianjin 300384,China)

机构地区：[1]华北理工大学电气工程学院,河北唐山063210 [2]天津理工大学电气工程与自动化学院,天津300384

出　　处：《重庆理工大学学报（自然科学）》2022年第9期179-186,共8页Journal of Chongqing University of Technology：Natural Science

基　　金：国家自然科学基金项目(61203343);河北省自然科学基金项目(F2018209289)。

摘　　要：基于Faster R-CNN(faster region based convolutional neural networks)模型对目标图像分类识别和边界回归的一般原理,在基础卷积网络、多尺度特征融合以及加入注意力机制和利用生物视觉的稀疏特性方面对Faster R-CNN网络模型做出改进。为验证所提出模型的性能,采用CTSDS(Chinese traffic sign dataset)数据集,在分析现实中影响交通标志检测性能因素的基础上,对数据集的部分样本做遮挡、加噪处理,实现数据增广。Matlab仿真结果表明:改进的网络模型对5类目标检测的平均精度为98.8%,证明了模型的有效性。Based on the general principle of Faster R-CNN(Faster Region based Convolutional Neural Networks)model classification of target images and boundary regression,the Faster R-CNN network model is improved through the basic convolutional network,multi-scale feature integration and the sparse characteristics of biological vision.In order to verify the performance of the proposed model,the CTSDS dataset(China Traffic Sign DataSet)is adopted.Simulation of this model through Matlab shows that the average accuracy of the improved network model for class 5 targets is 98.8%,which proves the effectiveness of the proposed model.

关 键 词：Faster R-CNN 数据增广 目标检测 

分 类 号：TP181[自动化与计算机技术—控制理论与控制工程]