[1]任汉俊,宋晓宁,於东军.一种新型粗-精表达策略行人检测方法[J].南京理工大学学报(自然科学版),2017,41(05):646.[doi:10.14177/j.cnki.32-1397n.2017.41.05.017]
　Ren Hanjun,Song Xiaoning,Yu Dongjun.Novel pedestrian detection method based on coarse-to-finerepresentation strategy[J].Journal of Nanjing University of Science and Technology,2017,41(05):646.[doi:10.14177/j.cnki.32-1397n.2017.41.05.017]
点击复制

一种新型粗-精表达策略行人检测方法()
分享到：

《南京理工大学学报》（自然科学版）[ISSN:1005-9830/CN:32-1397/N]

卷:
41卷
期数:
2017年05期
页码:
646
栏目:
出版日期:
2017-10-31

文章信息/Info

Title:
Novel pedestrian detection method based on coarse-to-finerepresentation strategy
文章编号:
1005-9830(2017)05-0646-07
作者:
任汉俊1; 宋晓宁1; 於东军2
1.江南大学 物联网工程学院,江苏 无锡 214122; 2.南京理工大学 计算机科学与工程学院,江苏 南京 210094
Author(s):
Ren Hanjun1; Song Xiaoning1; Yu Dongjun2
1.School of IoT Engineering,Jiangnan University,Wuxi 214122,China; 2.School of ComputerScience and Engineering,Nanjing University of Science and Technology,Nanjing 210094,China
关键词:
行人检测;  局部无关通道特征;  颜色自相似特征;  卷积网络结构;  平均对数漏检率
Keywords:
pedestrian detection;  locally decorrelated channel feature;  color self-similarity feature;  convolutional network architecture;  log-average miss rate
分类号:
TP391
DOI:
10.14177/j.cnki.32-1397n.2017.41.05.017
摘要:
为了遏制行人检测过程中产生的过多的误检窗口,该文在局部无关通道特征(LDCF)方法基础上提出了一种基于粗-精表达策略的新型行人检测方法。首先运用LDCF方法对行人进行粗略检测,产生一系列高召回率的候选窗口; 然后通过改进颜色自相似特征和引入简化的卷积网络结构,进一步提取这些窗口的鲁棒融合特征; 最后应用级联分类器对候选窗口进行精细分类判断。在行人检测数据集INRIA和Caltech上的实验结果表明,与传统的行人检测方法LDCF相比,该文方法的平均对数漏检率分别降低2.81%和3.85%,充分验证了该文策略的有效性和特征的鲁棒性。
Abstract:
In order to suppress too many false detection windows in pedestrian detection,a novel coarse-to-fine representation strategy is proposed based on the locally decorrelated channel features(LDCF)method.The LDCF method is used for coarse detection to generate a set of candidate windows with the high recall rate,and the improved color self-similarity feature extraction method and the simplified convolutional network architecture are introduced to extract more discriminant fusion features over these windows.Finally,the cascade classifier is applied for the fine classification of candidate windows.Experimental results show that,compared with the traditional LDCF method,the log-average miss rates of the proposed method on the INRIA and the Caltech databases are reduced by 2.81% and 3.85% respectively,demonstrating merits of the coarse-to-fine strategy and robustness of extracted features.

参考文献/References:

[1] 苏松志,李绍滋,陈淑媛,等.行人检测技术综述[J].电子学报,2012,40(4):814-820.
Su Songzhi,Li Shaozi,Chen Shuyuan,et al.A survey on pedestrian detection[J].Acta Electronica Sinica,2012,40(4):814-820.
[2]Dalal N,Triggs B.Histograms of oriented gradients for human detection[C]//IEEE Conference on Computer Vision and Pattern Recognition.San Diego,US:IEEE Computer Society Press,2005:886-893.
[3]Gan Guolong,Cheng Jian.Pedestrian detection based on HOG-LBP feature[C]//International Conference on Computational Intelligence and Security.Sanya,China:CIS,2011:1184-1187.
[4]Dollár P,Appel R,Belongie S,et al.Fast feature pyramids for object detection[J].IEEE Transactions on Pattern Analysis and Machine Intelligence,2014,36(8):1532-1545.
[5]Nam W,Dollár P,Han J H.Local decorrelation for improved pedestrian detection[J].Advances in Neural Information Processing System,2014(1):424-432.
[6]Zhang Liliang,Lin Liang,Liang Xiaodan,et al.Is faster R-CNN doing well for pedestrian detection?[C]//Europeon Conference on Computer Vision.Amsterdam,Netherlands:Springer International Publishing,2016:443-457.
[7]顾会建,陈俊周.基于改进颜色自相似特征的行人检测方法[J].计算机应用,2014,34(7):2033-2035.
Gu Huijian,Chen Junzhou.Pedestrian detection based on improved color self-similarity feature[J].Journal of Computer Applications,2014,34(7):2033-2035.
[8]Mu Yadong,Yan Shuicheng,Liu Yi,et al.Discriminative local binary patterns for human detection in personal album[C]//IEEE Conference on Computer Vision and Pattern Recognition.Anchorage,US:IEEE Computer Society Press,2008:1-8.
[9]Chan T H,Jia Kui,Gao Shenghua,et al.PCANet:A simple deep learning baseline for image classification[J].IEEE Transactions on Image Processing,2014,24(12):5017-5032.
[10]Ke Wei,Zhang Yao,Wei Pengxu,et al.Pedestrian detection via PCA filters based convolutional channel features[C]//IEEE International Conference on Acoustics,Speech and Signal Processing.Brisbane,Australia:IEEE Computer Society Press,2015:1394-1398.
[11]Dollár P,Wojek C,Schiele B,et al.Pedestrian detection:An evaluation of the state of the art[J].IEEE Transactions on Pattern Analysis and Machine Intelligence,2012,34(4):743-761.
[12]Benenson R,Mathias M,Tuytelaars T,el al.Seeking the strongest rigid detector[C]//IEEE Conference on Computer Vision and Pattern Recognition.Potland,US:IEEE Computer Society Press,2013:3666-3673.
[13]Lim J J,Zitnick C L,Dollár P.Sketch tokens:A learned mid-level representation for contour and object detection[C]//IEEE Conference on Computer Vision and Pattern Recognition.Potland,US:IEEE Computer Society Press,2013:3158-3165.
[14]Paisitkriangkrai S,Shen Chunhua,Hengel A V D.Strengthening the effectiveness of pedestrian detection[J].Lecture Notes in Computer Science,2014,8692:546-561.
[15]Benenson R,Omran M,Hosang J,et al.Ten years of pedestrian detection,what have we learned?[C]//Europeon Conference on Computer Vision.Eurich,Switzerland:Springer International Publishing,2014,8926:613-627.

相似文献/References:

[1]杨 阳,杨静宇.基于显著性分割的红外行人检测[J].南京理工大学学报(自然科学版),2013,37(02):251.
　Yang Yang,Yang Jingyu.Pedestrian detection of infrared images based on saliency segmentation[J].Journal of Nanjing University of Science and Technology,2013,37(05):251.
[2]周少康,宋晓宁,於东军.基于区域候选的粗-精行人检测方法[J].南京理工大学学报(自然科学版),2020,44(03):272.[doi:10.14177/j.cnki.32-1397n.2020.44.03.003]
　Zhou Shaokang,Song Xiaoning,Yu Dongjun.Coarse-to-fine method of pedestrian detection based on region proposal[J].Journal of Nanjing University of Science and Technology,2020,44(05):272.[doi:10.14177/j.cnki.32-1397n.2020.44.03.003]

备注/Memo

备注/Memo:
收稿日期:2016-12-23 修回日期:2017-04-17

基金项目:国家科技支撑计划(2015BAD17B02); 国家自然科学基金(61672265); 江苏省自然科学基金(BK20161135); 中国博士后科学基金(2016M590407)
作者简介:任汉俊(1993-),男,硕士生,主要研究方向:人工智能与模式识别,E-mail:hanjun_ren@163.com; 通讯作者:於东军(1975-),男,博士,教授,博士生导师,主要研究方向:模式识别与智能系统,E-mail:njyudj@njust.edu.cn。
引文格式:任汉俊,宋晓宁,於东军.一种新型粗-精表达策略行人检测方法[J].南京理工大学学报,2017,41(5):646-652.
投稿网址:http://zrxuebao.njust.edu.cn
更新日期/Last Update: 1900-01-01