联合CRF和可变部位模型的行人检测方法

系统仿真学报 ›› 2015, Vol. 27 ›› Issue (10): 2310-2315.

联合CRF和可变部位模型的行人检测方法

马技^1,2, 李晶皎¹, 马利^1,2, 赵越³

1.东北大学信息科学与工程学院,辽宁沈阳 110819;
2.辽宁大学信息学院,辽宁沈阳 110036;
3.渤海大学,辽宁锦州 121000

收稿日期:2015-06-14 修回日期:2015-07-23 出版日期:2015-10-08 发布日期:2020-08-07
作者简介:马技(1980-),男,山东陵县,博士生,讲师,研究方向为计算机视觉;李晶皎(1964-),女,辽宁沈阳,博士,教授,研究方向为计算机视觉;马利(1978-),女,辽宁黑山,博士生,讲师,研究方向为计算机视觉。
基金资助:
辽宁省教育厅一般科技项目(L2012003)

Combining CRF and Deformable Part Model for Pedestrian Detection

Ma Ji^1,2, Li Jingjiao¹, Ma Li^1,2, Zhao Yue³

1. School of Information Science & Engineering, Northeastern University, Shenyang 110819, China;
2. School of Information, Liaoning University, Shenyang 110036, China;
3. Bohai University, Jinzhou 121000, China

Received:2015-06-14 Revised:2015-07-23 Online:2015-10-08 Published:2020-08-07

摘要/Abstract

摘要： 行人目标检测在许多领域有着广泛的应用,它是计算机视觉研究的焦点之一。基于部位的检测方法在行人检测方面表现出非常出色的性能,在人体姿态变化方面具有很强的适应性,但是对于部位遮挡问题效果不佳。当判别阈值较高的时候,漏检率很高。考虑LSVM方法对遮挡信息挖掘不足,在可变部位模型的基础上,针对部位遮挡问题,建立了条件随机场模型,采用两层分类器。在参数学习中,采用随机梯度下降和置信传播算法优化条件随机场的目标函数。实验结果表明,该文提出的方法在遮挡问题方面表现出较好的效果。

关键词: 行人检测, 可变部位模型, 条件随机场, 隐支持向量机

Abstract: Pedestrian detection has been widely used in many fields. It is one of the focus in computer vision. The part-based detection method in the pedestrian detection shows excellent performance and has a strong adaptability in posture change of human body. But it is not good for Occlusion problem. When the Discriminative threshold is higher, miss rate is very high. Considering the disadvantage of LSVM method for mining hidden information, a two layers classifier was proposed based on the deformable parts model establishing conditional random field model for Occlusion problem. For learning model parameters, the stochastic gradient descent and belief propagation algorithm optimization objective function of the random field conditions were used. The experimental results show that the proposed approach achieves good effect for Occlusion problem.

Key words: pedestrian detection, deformable part model, CRF, LSVM

中图分类号:

TP391

马技, 李晶皎, 马利, 赵越. 联合CRF和可变部位模型的行人检测方法[J]. 系统仿真学报, 2015, 27(10): 2310-2315.

Ma Ji, Li Jingjiao, Ma Li, Zhao Yue. Combining CRF and Deformable Part Model for Pedestrian Detection[J]. Journal of System Simulation, 2015, 27(10): 2310-2315.

参考文献

[1] Xiaogang Wang, Meng Wang, Wei Li.Scene-Specific Pedestrian Detection for Static Video Surveillance[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence (S0162-8828), 2014, 36(2): 361-374.
[2] M Everingham, L Van Gool, J Winn, et al. The PASCAL Visual Object Classes (VOC) Challenge[J]. International Journal of Computer Vision (S0920-5691), 2010, 88(2): 303-338.
[3] P F Felzenszwalb, D McAllester, D Ramanan.Object detection with discriminatively trained part-based models[J]. IEEE Pattern Analysis and Machine Intelligence (S0162-8828), 2010, 32(9): 1627-1645.
[4] http://host.robots.ox.ac.uk:8080/pascal/VOC/
[DB/OL].(2014-11)[2015-06].
[5] P Ott, M Everingham.Shared parts for deformable part-based model[J]. Proceedings of IEEE Conference on Computer Vision and Pattern Recognition (S1063-6919), 2011, 6(1): 1513-1520.
[6] H T Niknejad, S Mita, D McAllester, et al. Vision based vehicle detection with discriminately trained mixture of weighted deformable part models[J]. IEEE Intelligent Transportation Systems (S2153-0009), 2011, 10(1): 1560-1565.
[7] L Zhu, Y Chen, A Yuille, et al. Latent hierarchical structural learning for object detection[J]. Proceedings of IEEE Conference on Computer Vision and Pattern Recognition (S1063-6919), 2010, 6(1): 1062-1069.
[8] M Enzweiler, A Eigenstetter, B Schiele, et al. Multi-cue pedestrian classification with partial occlusion handling[J]. Proceedings of IEEE Conference on Computer Vision and Pattern Recognition (S1063-6919), 2010, 6(1): 990-997.
[9] G Duan, H Ai, S Lao.A structural filter approach to human detection[J]. ECCV (S0302-9743), 2010, 9(1): 238-251.
[10] M Enzweiler, A Eigenstetter, B Schiele, et al. Gavrila. Multi-cue pedestrian classification with partial occlusion handling[C]// Proceedings of IEEE Conference on Computer Vision and Pattern Recognition (S1063-6919). USA: IEEE, 2010.
[11] T Wu, S Zhu.A numeric study of the bottom-up and top-down inference processes in and-or graphs[J]. Int’l Journal of Computer Vision (S0920-5691), 2011, 93(2): 226-252.
[12] W Fang, Y Zhao, Y Yuan K.Real time multiple vehicles tracking with occlusion handling[C]// Sixth International Conference on Image and Graphic (S978-0-7695-4541-7). USA: IEEE, 2011, 8: 667-672.
[13] H T Niknejad, K Takeuchi, S Mita, et al. On-road multi vehicle tracking using deformable object model and particle filter with improved likelihood estimation[J]. IEEE Transactions on Intelligent Transportation Systems (S1524-9050), 13(2): 748-758.
[14] W Ouyang, X Wang.A discriminative deep model for pedestrian detection and occlusion handling[J]. Proceedings of IEEE Conference on Computer Vision and Pattern Recognition (S1063-6919), 2012, 6(1): 3258-3265.
[15] S Sivaraman, M M Trivedi.Real-time vehicle detection using part at intersections[J]. Proc. IEEE Intelligent Transportation Systems (S2153-0009), 2012, 9(1): 1519-1524.
[16] A Quattoni, M Collins, T Darrell.Conditional random fields for object recognition[J]. NIPS (S0018-9448), 2004, 12: 1097-1104.
[17] N Dalal, B Triggs.Histograms of oriented gradients for human detection[J]. Proceedings of IEEE Conference on Computer Vision and Pattern Recognition (S1063-6919), 2005, 6(1): 886-893.
[18] P Felzenszwalb, D Huttenlocher.Distance transforms of sampled functions[M]. USA: Cornell University CIS, Tech. Rep. 2004: 2004-1963.
[19] S Andrews, I Tsochantaridis, T Hofmann.Support Vector Machines for Multiple-Instance Learning[J]. NIPS (S0018-9448), 2003, 12(1): 577-584.
[20] S Shalev-Shwartz, Y Singer, N Srebro.Pegasos: primal estimated sub-gradient solver for SVM[J]. Mathematical Programming (S0025-5610), 2011, 127(1): 3-30.