《计算机应用研究》|Application Research of Computers

一种改进的深度残差网络行人检测方法

Improved pedestrian detection method based on depth residual network

免费全文下载 (已被下载 次)  
获取PDF全文
作者 郝旭政,柴争义
机构 天津工业大学 计算机科学与软件学院,天津 300387
统计 摘要被查看 次,已被下载
文章编号 1001-3695(2019)05-060-1569-04
DOI 10.19734/j.issn.1001-3695.2017.12.0836
摘要 为了提高行人检测方法的准确率,针对行人图像特征,提出一种基于深度残差网络和YOLO(you only look once)方法的行人检测方法。以加强行人特征表达为目的,通过分析行人在图像中的表达和分布特征,提出一种不影响实时性的矩形输入深度残差网络分类模型以改进YOLO检测方法,使模型能够更好地表征行人;为了进一步提高模型的准确率和泛化能力,采用了混合行人数据集训练的方式,提取VOC数据集的行人数据与INRIA数据集组成混合数据集进行训练,明显降低了漏检率;并且利用聚类分析预测框的方法重新设计了初始预测框,提高行人定位能力并加快收敛。经公开的INRIA数据集的测试实验证明,该方法较主流的行人检测方法每张图片误检率有明显改善,降低至13.86%,有1.51%~58.62%不同程度的提升,并且本方法拥有良好的实时性和泛化能力,实用性强。
关键词 行人识别; 深度残差网络; YOLOv2; 卷积神经网络; 深度学习
基金项目 国家自然科学基金资助项目(U1504613)
本文URL http://www.arocmag.com/article/01-2019-05-060.html
英文标题 Improved pedestrian detection method based on depth residual network
作者英文名 Hao Xuzheng, Chai Zhengyi
机构英文名 School of Computer Science & Software Engineering,Tianjin Polytechnic University,Tianjin 300387,China
英文摘要 To improve the accuracy of the pedestrian detection method, this paper proposed a rectangular input of convolution neural network enhance the new pedestrian detection method based on the depth residual network and YOLO object detection method. The rectangular input helped the model gain the pedestrian characteristics expression by analyzing the expression and distribution characteristics of pedestrians in the images. The depth residual network with pre-activation for YOLO object detection improved the feature extraction ability through more layers of convolution neural networks. Hybrid dataset training and cluster anchor boxes could also improve the pedestrian detection performance. The test results of INRIA dataset prove that the method has better detection performance than the popular pedestrian detection methods, the index of false positive per image can reduce to 13.86%, improving ranging from 1.51% to 58.62% in varying degrees.
英文关键词 pedestrian detection; deep residual network; YOLOv2; convolutional neural network; deep learning
参考文献 查看稿件参考文献
 
收稿日期 2017/12/22
修回日期 2018/2/8
页码 1569-1572,1584
中图分类号 TP391.41
文献标志码 A