基于二维步态的身份识别
|
摘要
步态识别是指通过自动提取视频图像中行人的运动特征来识别人的身份,它具有非侵犯性、远距离识别、难于隐藏等特点。近年来,随着机场、银行、军事基地等安全敏感场合对大范围视觉监控系统的需求增加,步态识别被越来越多的研究者所关注。本文在总结和分析国内外相关工作的基础上,研究了二维步态识别方法,主要内容如下:
考虑到人体运动轮廓的时空相关特性,提出一种基于序列图像轮廓相似性的步态识别方法,首先在图像序列上使用背景减除法提取人体轮廓,然后截取下肢轮廓作为特征模板,最后计算序列图像中下肢轮廓之间的相似性实现目标的身份识别。由于个体步态周期不同步问题,论文引入动态时间规整算法,对行走过程中可能出现的速率变化进行非线性调整。试验结果表明这种方法的识别率令人满意,计算代价相对较低。
考虑到人体大腿运动是一种单摆模型,提出一种基于人体运动信息的步态识别方法,使用Canny检测方法和哈夫变换提取两大腿之间的角度之后,再融合人体轮廓的宽,度进行身份识别。在步态识别原型系统上对该方法进行验证,和其它方法相比,识别率明显提高。
Gait recognition aims essentially to recognize people by automatically extracting movement characteristic of walking person in the video. Gait recognition technology has its own advantages which is un-invasive, distance recognition and difficult to conceal. With an increasing demand for a full range of visual surveillance systems in security-sensitive environments such as banks and airports, Gait recognition has recently gained more and more interests from researchers. Based on analysis and summary of related work at home and abroad, two dimensional gait recognition methods are researched in this paper. The main content is as follows:
In consider of temporal and spatial correlation of human body contour, we propose a gait recognition method which is based on the similarity of image sequence contours. Firstly, a background subtraction method is used to extract the body contours in each image sequence. Secondly, we intercept the contours of the body lower limbs which are used as feature template. Finally, to recognize identity of object, we calculate similarities between the contours of lower limbs from image sequences. DTW (dynamic time warping) algorithm is introduced to adjust nonlinearly velocity change during walking because of asynchrony problem of gait cycle. Experimental results demonstrate that this method has an encouraging recognition rate with relatively low computational cost.
Considering that the upper leg of human body is a pendulum model, we propose a gait recognition method which is based on movement information of the body. Using Canny detection method and Hough transform, Angle between two upper legs is extracted in image sequence. Then, these angles are combined with widths of human contours to form the vector which is used to recognize identity. The proposed method is verified in prototype system and results demonstrate that recognition rate of this method is much higher than other method.
考虑到人体运动轮廓的时空相关特性,提出一种基于序列图像轮廓相似性的步态识别方法,首先在图像序列上使用背景减除法提取人体轮廓,然后截取下肢轮廓作为特征模板,最后计算序列图像中下肢轮廓之间的相似性实现目标的身份识别。由于个体步态周期不同步问题,论文引入动态时间规整算法,对行走过程中可能出现的速率变化进行非线性调整。试验结果表明这种方法的识别率令人满意,计算代价相对较低。
考虑到人体大腿运动是一种单摆模型,提出一种基于人体运动信息的步态识别方法,使用Canny检测方法和哈夫变换提取两大腿之间的角度之后,再融合人体轮廓的宽,度进行身份识别。在步态识别原型系统上对该方法进行验证,和其它方法相比,识别率明显提高。
Gait recognition aims essentially to recognize people by automatically extracting movement characteristic of walking person in the video. Gait recognition technology has its own advantages which is un-invasive, distance recognition and difficult to conceal. With an increasing demand for a full range of visual surveillance systems in security-sensitive environments such as banks and airports, Gait recognition has recently gained more and more interests from researchers. Based on analysis and summary of related work at home and abroad, two dimensional gait recognition methods are researched in this paper. The main content is as follows:
In consider of temporal and spatial correlation of human body contour, we propose a gait recognition method which is based on the similarity of image sequence contours. Firstly, a background subtraction method is used to extract the body contours in each image sequence. Secondly, we intercept the contours of the body lower limbs which are used as feature template. Finally, to recognize identity of object, we calculate similarities between the contours of lower limbs from image sequences. DTW (dynamic time warping) algorithm is introduced to adjust nonlinearly velocity change during walking because of asynchrony problem of gait cycle. Experimental results demonstrate that this method has an encouraging recognition rate with relatively low computational cost.
Considering that the upper leg of human body is a pendulum model, we propose a gait recognition method which is based on movement information of the body. Using Canny detection method and Hough transform, Angle between two upper legs is extracted in image sequence. Then, these angles are combined with widths of human contours to form the vector which is used to recognize identity. The proposed method is verified in prototype system and results demonstrate that recognition rate of this method is much higher than other method.
引文
[1] Niyogi S. A , Adelson E. H. Analyzing and recognizing walking figures in XYT[A]. Proceeding of CVPR1994 Conference[C]. Seattle, 1994:469-474
[2] Murase H , Sakai R. Moving object recognition in eigenspace representation: gait analysis and lip reading[J]. Pattern Recognition Letters, 1996, V17(2): 155-162
[3] Huang P. S, Harris C. J, Nixon M. S. Recognizing humans by gait via parametric canonical space[J].Artificial Intelligence in Engineering, 1996, V13( 4): 359-366
[4] Little J, Boyd J. Recognizing people by their gait: the shape of motion[J]. Videre, 1998, V14(6):83-105
[5] He Q , Debrunner C. Individual recognition from periodic activity using Hidden Markov Models[A].Proceeding of WHM2000 Conference[C]. Austin, 2000
[6] Kale A, Rajagopalan A, Cuntoor N et al. Gait-based recognition of humans using continuous HMMs[A].Proceeding of ICAFGR2002 Conference[C]. Washington, 2002:366-341
[7] BenAbdelkade C, Culter R, Nanda H, et al. EigenGait: motion-based recognition of people using image self-similarity[A]. Proceeding of AVBPA 2001[C]. Halmstad, 2001:284-294
[8] Hayfron-Acquah J, Nixon M, Carter J. Automatic gait recognition by symmetry analysis[J]. Pattern Recognition Letters, 2003, V24(13):2175-2183
[9] Lee L, Grimson W. E. L. Gait appearance for recognition [A]. ECCV Workshop on Biometric Authentication[C]. Copenhagen, 2002:143-154
[10] Sarkar S, Phillips P.J, Liu Z, et al. The humanID gait challenge problem: data sets, performances, and analysis[J]. Proceeding of PAMI Conference, 2005, V27(2):162-177
[11] Shutler J. D, Nixon M. S. Zernike Velocity Moments for Description and Recognition of Moving Shapes[A]. Proceeding of BMVC2001 Conference[C]. Manchester, 2001:705-714
[12] Robledo I, Sarkar S. Experiments on gait analysis by exploiting nonstationarity in the distribution of feature relationships[A]. Proceeding of ICPR Conference[C]. Quebec, 2002:1-4
[13] Collins R. T, Gross R, Shi J. Silhouette-based human identification from body Shape and gait[A].Proceeding of ICAFGR2002 Conference[C]. Washington, 2002:351-356
[14] Tolliver D, Collins R. Gait Shape Estimation for Identification[A]. Proceeding of ICAVBP2003 Conference[C]. UnivSurrey, Guildford, UK, 2003:734-742
[15] Foster J. P, Nixon M. S, Prugel-Bennet A. Automatic Gait Recognition using Area-Based Metrics[J].Pattern Recognition Letters, 2003, V24(14):2489-2497
[16] Bobick A.F, Davis J.W. The recognition of human movement using temporal templates[J].Transactions on Pattern Analysis and Machine Intelligence, 2001, V 23(3):257 - 267
[17] Han J, Bhanu B. Individual recognition using gait energy image[J]. Transactions on Pattern Analysis and Machine Intelligence, 2006, V28(2)
[18] Boyd J. Video phase-locked loops in gait recognition[A]. Proceeding of ICCV2001 Conference[C].Vancouver BC, 2001:696 -703
[19] Lie A. S, Enokida S, Wada T, et al. Magnitude and Phase Spectra of Foot Motion for Gait Recognition[A]. Proceeding of ICCAIP2005 Conference[C]. 2005:390-397
[20] Cunadom D, Nixon M, Carter J. Using gait as a biometric, via phase-weighted magnitude spectra[A]. Proceeding of ICAVBBPA1997 Conference[C]. Crans-Montana, 1997:95-102
[21] Cunado D, Nixon M, Carter J.N. Automatic extraction and description of human gait models for recognition purposes[J]. Computer Vision and Image Understanding, 2003, V90(1):1-41
[22] Yam C, Nixon M, Carter J. Gait recognition by walking and running: a model-based approach[A].Proceeding of ACCV2002 Conference[C]. Melbourne, 2002:1-6
[23] Yam C, Nixon M, Carter J.N. On the relationship of human walking and running automatic person identification by gait[A]. Proceeding of ICPR 2002 Conference[C]. 2002:287-290
[24] Yao J. H, Nixon M, Harris C. J. Extracting gait signatures based on anatomical knowledge[A].Proceeding of BMVA Symposium on Advancing Biometric Techniques at the Royal Statistical Society[C].London, 2002
[25] Yoo J.H, Nixon M.S, Harris C.J. Model-driven statistical analysis of human gait motion[A].Proceeding of ICIP2002 Conference[C]. 2002:285-288
[26] Zhang R, Vogler C, Metaxas D. Human gait recognition at sagittal plane[J]. Image and Vision Computing, 2007,V25(3):321-330
[27] Bobick A, Johnson A. Gait Recognition Using Static, Activity-Specific Parameters[A]. Proceeding of CVPC2001 Conference[C]. Kauai, Hawaii, 2001:423-430
[28] BenAbdelkader C, Cutler R.G, Davis L. S. Stride and cadence as a biometric in automatic person identification and verification[A]. Proceeding of ICAFGR2002 Conference[C]. Washington DC,2002:372-377
[29] BenAbdelkader C, Cutler R.G, Davis L.S. View invariant estimation of height and stride for gait recognition [A]. Proceedings of ICBA2002 Conference[C]. Copenhagen, 2002:155-167
[30] BenAbdelkader C, Cutler R.G, Davis L.S. Person identification using automatic height and stride estimation[A]. Proceeding of ICPR2002[C]. Quebec, 2002:377-380
[31] Tanawongsuwan R , Bobick A. Gait recognition from time-normalized joint-angle trajectories in the walking plane[A]. Proceeding of ICCVPR2001 Conference[C]. Kauai, Hawaii, 2001:726-731
[32] Bhanu B, Han J. Individual recognition by kniematic-based gait analysis[A]. Proceeding of ICPR2002 Conference[C]. Quebec, 2002:343-346
[33] Bhanu B, Han J. Bayesian based performance prediction for gait recognition[A]. Proceeding of WMVC2002 Conference[C]. Orlando, 2002:145-150
[34] Bhanu B, Han J. Performance prediction for individual recognition by gait[J]. Pattern Recognition Letters, 2005, V26(5):615-624
[35] Urtasun R, Fua P, 3D tracking for gait characterization and recognition[J]. Proceeding of ICAFGR2004 Conference, Seoul, 2004:17-22
[36] Zhao G, Liu G, Li H, et al. 3D Gait Recognition Using Multiple Cameras[A]. Proceeding of ICAFRGR2006 Conference[C]. Southampton, 2006:529-534
[37] Lee C. S, Elgammal A. Gait tracking and recognition using person-dependent dyanmic shape model[A]. Proceeding of ICAFGR2006 Conference[C]. Southampton, 2006:553-559
[38] Wang L, Tan T.N, Ning H.Z, et al. Fusion of static and dynamic body biometrics for gait recognition[J].Transactions on Circuits and Systems for Video, 2002, V14(12):149-158
[39] Bazin A, Nixon M. Probabilistic combination of static and dynamic gait features for verification[A]. Proceeding of BTHI2005 ConferencefC]. Orlando, 2005:23-30
[40] Wagg D.K, Nixon M.S. Automated markless extraction of walking people using deformable contour models[J].Computer Animation and Virtual Worlds,2004,V15(3):399-406
[41]Veres.G.V,Gordon.L,Carter.J.N.et al.What image information is important in silhouette-based gait recognition?[A].Proceedings of CSCCVPR[C].Washington DC,2004:776-782
[42]张宏林.Visual C++数字图像模式识别技术及工程实践[M].北京:人民邮电出版社.2003:299-312
[43]章毓晋.图像分析[M].北京:清华大学出版社.2005:139-150
[44]孙即祥.现代模式识别[M].2.长沙:国防科技大学出版社.2003:172-189
[2] Murase H , Sakai R. Moving object recognition in eigenspace representation: gait analysis and lip reading[J]. Pattern Recognition Letters, 1996, V17(2): 155-162
[3] Huang P. S, Harris C. J, Nixon M. S. Recognizing humans by gait via parametric canonical space[J].Artificial Intelligence in Engineering, 1996, V13( 4): 359-366
[4] Little J, Boyd J. Recognizing people by their gait: the shape of motion[J]. Videre, 1998, V14(6):83-105
[5] He Q , Debrunner C. Individual recognition from periodic activity using Hidden Markov Models[A].Proceeding of WHM2000 Conference[C]. Austin, 2000
[6] Kale A, Rajagopalan A, Cuntoor N et al. Gait-based recognition of humans using continuous HMMs[A].Proceeding of ICAFGR2002 Conference[C]. Washington, 2002:366-341
[7] BenAbdelkade C, Culter R, Nanda H, et al. EigenGait: motion-based recognition of people using image self-similarity[A]. Proceeding of AVBPA 2001[C]. Halmstad, 2001:284-294
[8] Hayfron-Acquah J, Nixon M, Carter J. Automatic gait recognition by symmetry analysis[J]. Pattern Recognition Letters, 2003, V24(13):2175-2183
[9] Lee L, Grimson W. E. L. Gait appearance for recognition [A]. ECCV Workshop on Biometric Authentication[C]. Copenhagen, 2002:143-154
[10] Sarkar S, Phillips P.J, Liu Z, et al. The humanID gait challenge problem: data sets, performances, and analysis[J]. Proceeding of PAMI Conference, 2005, V27(2):162-177
[11] Shutler J. D, Nixon M. S. Zernike Velocity Moments for Description and Recognition of Moving Shapes[A]. Proceeding of BMVC2001 Conference[C]. Manchester, 2001:705-714
[12] Robledo I, Sarkar S. Experiments on gait analysis by exploiting nonstationarity in the distribution of feature relationships[A]. Proceeding of ICPR Conference[C]. Quebec, 2002:1-4
[13] Collins R. T, Gross R, Shi J. Silhouette-based human identification from body Shape and gait[A].Proceeding of ICAFGR2002 Conference[C]. Washington, 2002:351-356
[14] Tolliver D, Collins R. Gait Shape Estimation for Identification[A]. Proceeding of ICAVBP2003 Conference[C]. UnivSurrey, Guildford, UK, 2003:734-742
[15] Foster J. P, Nixon M. S, Prugel-Bennet A. Automatic Gait Recognition using Area-Based Metrics[J].Pattern Recognition Letters, 2003, V24(14):2489-2497
[16] Bobick A.F, Davis J.W. The recognition of human movement using temporal templates[J].Transactions on Pattern Analysis and Machine Intelligence, 2001, V 23(3):257 - 267
[17] Han J, Bhanu B. Individual recognition using gait energy image[J]. Transactions on Pattern Analysis and Machine Intelligence, 2006, V28(2)
[18] Boyd J. Video phase-locked loops in gait recognition[A]. Proceeding of ICCV2001 Conference[C].Vancouver BC, 2001:696 -703
[19] Lie A. S, Enokida S, Wada T, et al. Magnitude and Phase Spectra of Foot Motion for Gait Recognition[A]. Proceeding of ICCAIP2005 Conference[C]. 2005:390-397
[20] Cunadom D, Nixon M, Carter J. Using gait as a biometric, via phase-weighted magnitude spectra[A]. Proceeding of ICAVBBPA1997 Conference[C]. Crans-Montana, 1997:95-102
[21] Cunado D, Nixon M, Carter J.N. Automatic extraction and description of human gait models for recognition purposes[J]. Computer Vision and Image Understanding, 2003, V90(1):1-41
[22] Yam C, Nixon M, Carter J. Gait recognition by walking and running: a model-based approach[A].Proceeding of ACCV2002 Conference[C]. Melbourne, 2002:1-6
[23] Yam C, Nixon M, Carter J.N. On the relationship of human walking and running automatic person identification by gait[A]. Proceeding of ICPR 2002 Conference[C]. 2002:287-290
[24] Yao J. H, Nixon M, Harris C. J. Extracting gait signatures based on anatomical knowledge[A].Proceeding of BMVA Symposium on Advancing Biometric Techniques at the Royal Statistical Society[C].London, 2002
[25] Yoo J.H, Nixon M.S, Harris C.J. Model-driven statistical analysis of human gait motion[A].Proceeding of ICIP2002 Conference[C]. 2002:285-288
[26] Zhang R, Vogler C, Metaxas D. Human gait recognition at sagittal plane[J]. Image and Vision Computing, 2007,V25(3):321-330
[27] Bobick A, Johnson A. Gait Recognition Using Static, Activity-Specific Parameters[A]. Proceeding of CVPC2001 Conference[C]. Kauai, Hawaii, 2001:423-430
[28] BenAbdelkader C, Cutler R.G, Davis L. S. Stride and cadence as a biometric in automatic person identification and verification[A]. Proceeding of ICAFGR2002 Conference[C]. Washington DC,2002:372-377
[29] BenAbdelkader C, Cutler R.G, Davis L.S. View invariant estimation of height and stride for gait recognition [A]. Proceedings of ICBA2002 Conference[C]. Copenhagen, 2002:155-167
[30] BenAbdelkader C, Cutler R.G, Davis L.S. Person identification using automatic height and stride estimation[A]. Proceeding of ICPR2002[C]. Quebec, 2002:377-380
[31] Tanawongsuwan R , Bobick A. Gait recognition from time-normalized joint-angle trajectories in the walking plane[A]. Proceeding of ICCVPR2001 Conference[C]. Kauai, Hawaii, 2001:726-731
[32] Bhanu B, Han J. Individual recognition by kniematic-based gait analysis[A]. Proceeding of ICPR2002 Conference[C]. Quebec, 2002:343-346
[33] Bhanu B, Han J. Bayesian based performance prediction for gait recognition[A]. Proceeding of WMVC2002 Conference[C]. Orlando, 2002:145-150
[34] Bhanu B, Han J. Performance prediction for individual recognition by gait[J]. Pattern Recognition Letters, 2005, V26(5):615-624
[35] Urtasun R, Fua P, 3D tracking for gait characterization and recognition[J]. Proceeding of ICAFGR2004 Conference, Seoul, 2004:17-22
[36] Zhao G, Liu G, Li H, et al. 3D Gait Recognition Using Multiple Cameras[A]. Proceeding of ICAFRGR2006 Conference[C]. Southampton, 2006:529-534
[37] Lee C. S, Elgammal A. Gait tracking and recognition using person-dependent dyanmic shape model[A]. Proceeding of ICAFGR2006 Conference[C]. Southampton, 2006:553-559
[38] Wang L, Tan T.N, Ning H.Z, et al. Fusion of static and dynamic body biometrics for gait recognition[J].Transactions on Circuits and Systems for Video, 2002, V14(12):149-158
[39] Bazin A, Nixon M. Probabilistic combination of static and dynamic gait features for verification[A]. Proceeding of BTHI2005 ConferencefC]. Orlando, 2005:23-30
[40] Wagg D.K, Nixon M.S. Automated markless extraction of walking people using deformable contour models[J].Computer Animation and Virtual Worlds,2004,V15(3):399-406
[41]Veres.G.V,Gordon.L,Carter.J.N.et al.What image information is important in silhouette-based gait recognition?[A].Proceedings of CSCCVPR[C].Washington DC,2004:776-782
[42]张宏林.Visual C++数字图像模式识别技术及工程实践[M].北京:人民邮电出版社.2003:299-312
[43]章毓晋.图像分析[M].北京:清华大学出版社.2005:139-150
[44]孙即祥.现代模式识别[M].2.长沙:国防科技大学出版社.2003:172-189