|本期目录/Table of Contents|

[1]谢尧芳,苏松志,李绍滋*.基于稀疏编码的迁移学习及其在行人检测中的应用[J].厦门大学学报(自然科学版),2010,49(02):186.
 XIE Yao fang,SU Song zhi,LI Shao zi*.Transfer Learning Based on Sparse Coding and It′s Application in Pedestrian Detection[J].Journal of Xiamen University(Natural Science),2010,49(02):186.
点击复制

基于稀疏编码的迁移学习及其在行人检测中的应用(PDF)
分享到:

《厦门大学学报(自然科学版)》[ISSN:0438-0479/CN:35-1070/N]

卷:
49卷
期数:
2010年02期
页码:
186
栏目:
研究论文
出版日期:
2010-03-20

文章信息/Info

Title:
Transfer Learning Based on Sparse Coding and It′s Application in Pedestrian Detection
作者:
谢尧芳苏松志李绍滋*
厦门大学信息科学与技术学院,福建 厦门 361005
Author(s):
XIE YaofangSU SongzhiLI Shaozi*
School of Information Science and Technology,Xiamen University,Xiamen 361005,China
关键词:
行人检测行人分类迁移学习稀疏编码
Keywords:
pedestrian detection pedestrian classification transfer learning sparse coding
分类号:
TP 391.41
文献标志码:
-
摘要:
行人检测是计算机视觉领域中的研究热点,其实质是一个二分类问题.目前基于统计的行人检测技术已取得了一定进展,但大都需要大量的训练数据.针对这一问题,提出了一种基于迁移学习的半监督行人分类方法:首先基于稀疏编码,从任意的未标记样本中,学习到一个紧凑、有效的特征表示;然后通过迁移学习,将学习到的特征表示方法迁移到行人分类中.在MIT行人数据库上的实验结果表明:该方法能有效地刻画出行人的特征,提高行人分类的性能,在标记样本少的情况下仍具有良好的分类效果,因此可应用于行人检测中.
Abstract:
Pedestrian detection has become one of the hottest topics in the domain of computer vision. It can be considered as a twoclassification problem. Though it has achieved some progress, most of previous methods need lots of training data. This paper proposed a novel semisupervised method for pedestrian classification, which was based on transfer learning and just needed only a few labeled data. Firstly, we used sparse coding to learn a slightly higherlevel, more succinct feature representation from the unlabeled data that randomly downloaded from the Internet. Then we applied this representation to the target classification problem by transfer learning. When using MIT pedestrian database for experiments, the result demonstrates that our approach can improve the classification rate and even shows great power when the labeled data is scarce. Therefore, our approach has the application prospect in pedestrian detection. No matter how the application condition changes, it just needs a small number of labeled data, which is much fewer than those of previous.

参考文献/References:


[1]贾慧星,章毓晋.车辆辅助驾驶系统中基于计算机视觉的行人检测研究综述[J].自动化学报,2007,33(1):8490.
[2]朱文佳.基于机器学习的行人检测关键技术研究[D].上海:上海交通大学,2008.
[3]Oren M,Papageorgiou C,Sinha P,et al.Pedestrian detection using wavelet templates[C]//Proceedings of IEEE Conference on Computer Vision and Pattern Recognition.San Juan,Puerto Rico:IEEE,1997:193199.
[4]Viola P,Jones M,Snow D.Detecting pedestrians using patterns of motion and appearance[C]//Proceedings of 9th IEEE International Conference on Computer Vision. Paris,France:IEEE,2003:734741.
[5]Zhao L,Thorpe C.Stereo and neural networkbased pedestrian detection[J].IEEE Transactions on Intelligent Transportation Systems,2000,1(3):148154.
[6]Lowe D G.Distinctive image features from scaleinvariant keypoints[J].International Journal of Computer Vision,2004,60(2):91110.
[7]Shashua A,Gdalyahu Y,Hayun G.Pedestrian detection for driving assistance systems:single frame classification and system level performance[C]//Proceedings of IEEE Intelligent Vehicles Symposium.Parma,Italy:IEEE,2004:16.
[8]Mikolajczyk K,Schmid C,Zisserman A.Human detection based on a probabilistic assembly of robust part detectors[C]//Proceedings of European Conference on Computer Vision.Prague,Czech Republic:Springer,2004:6982.
[9]Dalal N,Triggs B.Histograms of oriented gradients for human detection[C]//Proceedings of IEEE Computer Society Conference on Computer Vision and Pattern Recognition.Washington DC,USA:IEEE,2005:886893.
[10]Wu B,Nevatia R.Detection of multiple,partially occluded humans in a single image by Bayesian combination of edgelet part detectors[C]//Proceedings of IEEE International Conference on Computer Vision.Washington DC,USA:IEEE,2005:9097.
[11]Gavrila D M,Philomin V.Realtime object detection for smart vehicles[C]//Proceedings of IEEE International Conference on Computer Vision.Fort Collins,Colorado,USA:IEEE,1999:8793.
[12]Sabzmeydani P,Mori G.Detecting pedestrians by learning shapelet features[C]//Proceedings of IEEE Computer Society Conference on Computer Vision and Pattern Recognition.Minneapolis,MN:IEEE,2007:18.
[13]Sinno J P,Qiang Y.A survey on transfer learning[OL].
[20081112].Available online:http://www.cse.ust.hk/~sinnopan/publications/TLsurvey\0822.pdf.
[14]Raina R,Andrew Y Ng,Koller D.Constructing informative priors using transfer learning[C]//Proceedings of the 23rd International Conference on Machine Learning.Pittsburgh,Pennsylvania,USA:ACM,2006:713720.
[15]Blitzer J,McDonald R,Pereira F.Domain adaptation with structural correspondence learning[C]//Proceedings of the Conference on Empirical Methods in Natural Language.Sydney:Morristown,2006:120128.
[16]Arnold A,Nallapati R,Cohen W W.A comparative study of methods for transductive transfer learning[C]//Proceedings of the 7th IEEE International Conference on Data Mining Workshops.Washington,DC,USA:IEEE,2007:7782.
[17]Raina R,Battle A,Lee H,et al.Selftaught learning:transfer learning from unlabeled data[C]//Proceedings of the 24th International Conference on Machine Learning.Corvalis,Oregon,USA:IEEE,2007:759766.
[18]Olshausen B A,Field D J.Sparse coding with an overcomplete basis set:a strategy employed by V1?[J]. Vision Research,1997,37:33113325.
[19]Lee H,Battle A,Raina R,et al.Efficient sparse coding algorithms[C]//Proceedings of Neural Information Processing Systems Conference.Cambridge,MA:MIT Press,2007:801808.

备注/Memo

备注/Memo:
收稿日期:20090508基金项目:国家自然科学基金(60873179);深圳市科技计划基础研究项目(JC200903180630A)*通讯作者:szlig@xmu.edu.cn
更新日期/Last Update: 2010-03-20