正在加载图片...
第14卷第2期 智能系统学报 Vol.14 No.2 2019年3月 CAAI Transactions on Intelligent Systems Mar.2019 D0:10.11992/tis.201710019 网络出版地址:http:/kns.cnki.net/kcms/detail/23.1538.TP.20180416.1401.010html 多层卷积特征的真实场景下行人检测研究 伍鹏瑛2,张建明2,彭建2,陆朝铨2 (1.长沙理工大学综合交通运输大数据智能处理湖南省重点实验室,湖南长沙410114;2.长沙理工大学计算 机与通信工程学院,湖南长沙410114) 摘要:针对真实场景下的行人检测方法存在漏检、误检率高,以及小尺寸目标检测精度低等问题,提出了一 种基于改进SSD网络的行人检测模型(PDIS)。PDIS通过引出更底层的输出特征图改进了原始SSD网络模型, 并采用卷积神经网络不同层输出的抽象特征对行人目标分别做检测,融合多层检测结果,提升了小目标行人的 检测性能。此外,针对数据集样本多样性能有效地提升检测算法的泛化能力,本文采集了不同光照、姿态、遮 挡等复杂场景下的行人图像,对背景比较复杂的NRIA行人数据集进行了扩充,在扩增的行人数据集上训练 的PDIS模型,提高了在真实场景下的行人检测精度。实验表明:PDIS在INRIA测试集上测试结果达到 93.8%的准确率,漏检率低至7.4%。 关键词:行人检测:卷积神经网络:SSD:真实场景:多尺度特征:目标检测:小目标行人:行人数据集 中图分类号:TP391文献标志码:A文章编号:1673-4785(2019)02-0306-10 中文引用格式:伍鹏瑛,张建明,彭建,等.多层卷积特征的真实场景下行人检测研究{J.智能系统学报,2019,14(2): 306-315. 英文引用格式:VU Pengying,.ZHANGJianming,.PENGJian,,etal.Research on pedestrian detection based on multi--layer convolu- tion feature in real sceneJI.CAAI transactions on intelligent systems,2019,14(2):306-315. Research on pedestrian detection based on multi-layer convolution feature in real scene WU Pengying,ZHANG Jianming,PENG Jian'2,LU Chaoquan'2 (1.Hunan Provincial Key Laboratory of Intelligent Processing of Big Data on Transportation,Changsha University of Science and Technology,Changsha 410114,China;2.School of Computer and Communication Engineering,Changsha University of Science and Technology,Changsha 410114,China) Abstract:Pedestrian detection methods in real scenes face some problems due to the high miss detection and false de- tection as well as the low detection accuracy of small size objects.To solve these problems,a pedestrian detection mod- el based on improved SSD(PDIS)is proposed.The PDIS method improves the original SSD network model by extract- ing the lower-level output feature maps.It employs the abstract features of different convolutional neural network layers to detect pedestrians respectively,and then integrates the detection results of multi layers to increase the pedestrian de- tection performance for small sizes.Considering that the diversity of dataset can effectively enhance the generalization ability of detection algorithm,the paper expands the INRIA pedestrian dataset with complex background by collecting pedestrian images with different illumination,pose and occlusion.The PDIS method trained on expanded pedestrian dataset increases the precision rate of pedestrian detection in real scenes.The experiment results on INRIA test set indic- ate that the precision rate of PDIS algorithm is up to 93.8%and the miss rate is as low as 7.4%. Keywords:pedestrian detection:CNN:single shot multibox detector:real scene:multi-scale features:object detection: small target pedestrians;Pedestrian dataset 收稿日期:2017-10-31.网络出版日期:2018-04-16 行人检测是判断输入的图像或视频中是否含 基金项目:国家自然科学基金项目(61402053):湖南省教育厅 科研重点项目(16A008):湖南省交通厅科技项目 有行人,并准确的找出行人的具体位置。行人检 (201446):长沙理工大学研究生科研创新项目 (CX20I7SS19):长沙理工大学研究生课程建设项目 测作为目标检测的一个子方向,在视频监控、行 (KC201611). 通信作者:张建明.E-mail:jmzhang@csust..edu.cn. 人识别山、图像检索以及先进的驾驶员辅助系统DOI: 10.11992/tis.201710019 网络出版地址: http://kns.cnki.net/kcms/detail/23.1538.TP.20180416.1401.010.html 多层卷积特征的真实场景下行人检测研究 伍鹏瑛1,2,张建明1,2,彭建1,2,陆朝铨1,2 (1. 长沙理工大学 综合交通运输大数据智能处理湖南省重点实验室,湖南 长沙 410114; 2. 长沙理工大学 计算 机与通信工程学院,湖南 长沙 410114) 摘 要:针对真实场景下的行人检测方法存在漏检、误检率高,以及小尺寸目标检测精度低等问题,提出了一 种基于改进 SSD 网络的行人检测模型 (PDIS)。PDIS 通过引出更底层的输出特征图改进了原始 SSD 网络模型, 并采用卷积神经网络不同层输出的抽象特征对行人目标分别做检测,融合多层检测结果,提升了小目标行人的 检测性能。此外,针对数据集样本多样性能有效地提升检测算法的泛化能力,本文采集了不同光照、姿态、遮 挡等复杂场景下的行人图像,对背景比较复杂的 INRIA 行人数据集进行了扩充,在扩增的行人数据集上训练 的 PDIS 模型,提高了在真实场景下的行人检测精度。实验表明:PDIS 在 INRIA 测试集上测试结果达到 93.8% 的准确率,漏检率低至 7.4%。 关键词:行人检测;卷积神经网络;SSD;真实场景;多尺度特征;目标检测;小目标行人;行人数据集 中图分类号:TP391 文献标志码:A 文章编号:1673−4785(2019)02−0306−10 中文引用格式:伍鹏瑛, 张建明, 彭建, 等. 多层卷积特征的真实场景下行人检测研究[J]. 智能系统学报, 2019, 14(2): 306–315. 英文引用格式:WU Pengying, ZHANG Jianming, PENG Jian, et al. Research on pedestrian detection based on multi-layer convolu￾tion feature in real scene[J]. CAAI transactions on intelligent systems, 2019, 14(2): 306–315. Research on pedestrian detection based on multi-layer convolution feature in real scene WU Pengying1,2 ,ZHANG Jianming1,2 ,PENG Jian1,2 ,LU Chaoquan1,2 (1. Hunan Provincial Key Laboratory of Intelligent Processing of Big Data on Transportation, Changsha University of Science and Technology, Changsha 410114, China; 2. School of Computer and Communication Engineering, Changsha University of Science and Technology, Changsha 410114, China) Abstract: Pedestrian detection methods in real scenes face some problems due to the high miss detection and false de￾tection as well as the low detection accuracy of small size objects. To solve these problems, a pedestrian detection mod￾el based on improved SSD (PDIS) is proposed. The PDIS method improves the original SSD network model by extract￾ing the lower-level output feature maps. It employs the abstract features of different convolutional neural network layers to detect pedestrians respectively, and then integrates the detection results of multi layers to increase the pedestrian de￾tection performance for small sizes. Considering that the diversity of dataset can effectively enhance the generalization ability of detection algorithm, the paper expands the INRIA pedestrian dataset with complex background by collecting pedestrian images with different illumination, pose and occlusion. The PDIS method trained on expanded pedestrian dataset increases the precision rate of pedestrian detection in real scenes. The experiment results on INRIA test set indic￾ate that the precision rate of PDIS algorithm is up to 93.8% and the miss rate is as low as 7.4%. Keywords: pedestrian detection; CNN; single shot multibox detector; real scene; multi-scale features; object detection; small target pedestrians; Pedestrian dataset 行人检测是判断输入的图像或视频中是否含 有行人,并准确的找出行人的具体位置。行人检 测作为目标检测的一个子方向,在视频监控、行 人识别[1] 、图像检索以及先进的驾驶员辅助系统 收稿日期:2017−10−31. 网络出版日期:2018−04−16. 基金项目:国家自然科学基金项目 (61402053);湖南省教育厅 科研重点项目 (16A008);湖南省交通厅科技项目 (201446) ;长沙理工大学研究生科研创新项 目 (CX2017SS19);长沙理工大学研究生课程建设项目 (KC201611). 通信作者:张建明. E-mail:jmzhang@csust.edu.cn. 第 14 卷第 2 期 智 能 系 统 学 报 Vol.14 No.2 2019 年 3 月 CAAI Transactions on Intelligent Systems Mar. 2019
向下翻页>>
©2008-现在 cucdc.com 高等教育资讯网 版权所有