37 1 Vol. 37, No. 1 2011 1 ACTA AUTOMATICA SINICA January, 2011 1, 2 1, 2 1, 2,.,. ( ) ( );., ( ).,. DOI,,,, 10.3724/SP.J.1004.2011.00028 Human Identification Based on 3D Tracking Trajectory of Head Vertex JIA Li-Hao 1, 2 ZOU Jian-Hua 1, 2 CHE Kai 1, 2 Abstract A novel biometric method for human identification is proposed based on 3D tracking trajectory of head vertex, accompanied by a systematic study on the related basic problems. Vertical displacement in sagittal plane and lateral displacement in transverse plane can be extracted from the 3D tracking trajectory of head vertex. Previous work has demonstrated effective use of height parameters (height mean and height amplitude) and stride parameters (stride length and cadence) extracted from vertical displacement for human identification. In this paper, we further extract swing parameters (swing mean, swing amplitude, and swing angle) from lateral displacement as additional discriminant features. A group of discriminant and robust features are obtained by integrating height parameters, stride parameters, and swing parameters. Experimental results confirm the effectiveness of the proposed method. Key words identification Biometric trait, 3D tracking trajectory of head vertex, vertical displacement, lateral displacement, human (Biometrics), [1].,,, [2].,,, [3]., [4 5]., (Sagittal) (Transverse) (Coronal) [5 6],.,, 2010-03-02 2010-08-20 Manuscript received March 2, 2010; accepted August 20, 2010 (50177025) Supported by National Natural Science Foundation of China (50177025) 1. 710049 2. 710049 1. Systems Engineering Institute, Xi an Jiaotong University, Xi an 710049 2. State Key Laboratory for Manufacturing Systems Engineering, Xi an Jiaotong University, Xi an 710049,, [7 10] ;,,.,, [4, 6].,,,.,.,. 1
1 : 29, ;,. 1., ;,,, ;, ( ) ( ), ( ) ;, (Support vector machine, SVM)., ( ) ( ) [11].,,,. 4.. 5. L-K [15 16]. 6. 3 5,. 2 Fig. 2 Schematic diagram of experiment platform of human identification based on 3D tracking trajectory 3 (a) ( : 1 690 mm),. 3 (b),, 3 (b),. 1 Fig. 1 Schematic diagram of human identification based on 3D tracking trajectory of head vertex 2,, 2. 40. : 1.. X W Y W Z W, 2. 2., [12]., δ L-K [13]. 3. LoG (Laplacian of Gaussian) [14]. (a) (a) Stereo correspondence matching and tracking trajectory of head vertex in binocular images (b) (b) 3D tracking trajectory of head vertex 3 Fig. 3 Generation of 3D tracking trajectory of head vertex 3,,
30 37,. [17], ; [18] ;, [19] [20]. 3, X W X W,,. 3.1, ;,.. X W,,,, 4.,, 4,, ;.,, 5. 5 : 1),, ; 2), ; 3),,,. 4 Fig. 4 5 Fig. 5 Time-series curves of vertical displacement, vertical projection, and lateral projection Correspondences between curves of vertical and lateral projections and human gait 3.2,. 1). 6, P tr1, P tr2 P tr3 P tr1 P tr2 P tr2 P tr3 ( Z 0 = 1 640 mm ) p tr1 p tr2 p tr2 p tr3., X i Y i Z i, X i O i Y i. 2). X i Y i Z i X i O i Y i,, 6 P i
1 : 31 p i p tr1 p tr2 r i. α. 3)., 6 p tr1 p tr2 p tr2 p tr3 β., S = {(i, y i )} n., n (a) (a) Troughs on time-series curve of vertical displacement Fig. 6 6 Schematic diagram of local directions of motion and coordinates based on human body,,,. 7, : 1.. 7 (a) p tr1, p tr2, p tr3, p tr4, p tr5,. 2.. 7 (b) X i Y i Z i. 3.,, 7 (c). 7 (d).,,., 40. 4. T = {(x i, y i, z i )} n, H = {(i, z i )} n Fig. 7 (b) (b) Human body coordinates on curve of lateral projection (c) (c) Normalized 3D tracking trajectory (d) (d) Curves of vertical and lateral displacements 7 Diagram of normalized 3D tracking trajectories
32 37 4.1 µ H α H L stride f stride ( T stride ) V. 1) µ H ( : mm) α H ( : mm) µ H α H V. ( ) 4π H(i) = µ H + α H cos N i + ϕ H + ε H, i = 0, 1, 2,, n, N, ϕ H, ε H. µ H α H, µ H = 1 n z i, α H = A 2 H n + B2 H, A H = 2 n B H = 2 n n ( ) 4π (z i µ H ) cos N i n ( ) 4π (z i µ H ) sin N i 2) f stride ( : /s) ( T stride ( : s)) T stride, 5. T stride., T stride. f stride T stride, f stride = 1/T stride ( /s). 3) L stride ( : mm) L stride, 5.,. 4) V ( : mm/s) V = (f s D)/n, D ( : mm), f s ( : Hz). 4.2 µ S α S β S, µ S α S. β S. 1) µ S ( : mm) α S ( : mm), ( ) 2π S(i) = µ S + α S cos N i + ϕ S + ε S, i = 0, 1, 2,, n, ϕ S, ε S., µ S α S, µ S = 1 n y i, α S = A 2 S n + B2 S, A S = 2 n B S = 2 n n n ( ) 2π (y i µ S ) cos N i ( ) 2π (y i µ S ) sin N i 2) β S ( : ),, β S.,. 5 SVM. [11],, 1., HMS, HPS, SPS, HFS1 HFS2 {µ H } {µ H, α H } {L stride, f stride } 1 {µ H, α H, L stride, V } 2 {µ H, α H, L stride, f stride }; SFS, SSPS1 SSPS2 {µ S, α S, β S } 1 {µ S, α S, β S, L stride, V } 2 {µ S, α S, β S, L stride, f stride }; HSS1 HSS2
1 : 33 1 {µ H, α H, L stride, V, µ S, α S, β S } 2 {µ H, α H, L stride, f stride, µ S, α S, β S }. V = L stride f stride,, V f stride, HFS1, SSPS1 HSS1. 1 SVM.,. SVM LIBSVM [21], (Radical basis function, RBF), γ C. SVM Leave-One-Out Bootstrapping, : 1., ; 2., ; 3.,, ; 4. 50,. 6 6.1 2, 8, 58, 60 /, 752 480. 46. 40,, 6, 18, 828. Table 1 1, 828 828.,,,. 6.2 ; ( ) ( ) ( ) ;,,.,,.,. α i α 0 β i β 0,. α 0 β 0, α 0 = 4.0 mm, β 0 = 13.0.,. 823,, 8. 6.3,,, 46 25 35 45, 10, 10. 5, 9. Table of feature combinations HMS {µ H} SFS {µ S, α S, β S} HSS1 {µ H, α H, L stride, V, µ S, α S, β S} HPS {µ H, α H} SSPS1 {µ S, α S, β S, L stride, V } HSS2 {µ H, α H, L stride, f stride, µ S, α S, β S} SPS {L stride, f stride} SSPS2 {µ S, α S, β S, L stride, f stride} HFS1 {µ H, α H, L stride,v } HFS2 {µ H, α H, L stride, f stride}
34 37 Fig. 8 8 Histograms of features extracted from dataset (a) 25 (a) Sub-datasets containing 25 subjects (b) 35 (b) Sub-datasets containing 35 subjects (c) 45 (c) Sub-datasets containing 45 subjects (d) (d) Average recognition rates 9 Fig. 9 Recognition rates of sub-datasets with different feature combinations
1 : 35 Table 2 2 Average recognition rates of sub-datasets with different feature combinations (%) (%) (%) 25 35 45 25 35 45 25 35 45 HMS 46.0 41.0 32.5 SFS 32.3 25.1 21.0 HSS1 92.9 90.3 88.9 HPS 66.7 59.8 52.4 SSPS1 61.2 54.7 50.3 HSS2 92.6 90.2 88.9 SPS 27.9 21.9 18.1 SSPS2 64.8 57.6 53.8 HFS1 88.0 84.7 83.1 HFS2 88.4 85.9 84.3 9 (a) 9 (c),, HSS1 HSS2,, ;,. 9 (d) 25 35 45, 10,,. 9 (d),,, HSS1 HSS2,. 2. 9 2, [11] HFS1 HFS2 ; SFS, SSPS1 SSPS2 SPS, SFS ; HSS1 HSS2 HFS1 HFS2 4.9 %.,,,,., HSS1 HSS2 46, 50 88.0 % 89.5 %. 7.,,,,, ( ),. References 1 Jain A K, Flynn P, Ross A A. Handbook of Biometrics. New York: Springer-Verlag, 2007. 1 22 2 Li S Z, Schouten B, Tistarelli M. Biometrics at a distance: issues, challenges, and prospects. Handbook of Remote Biometrics for Surveillance and Security. London: Springer, 2009. 3 21 3 Nixon M S, Tan T N, Chellappa R. Human Identification Based on Gait. New York: Springer-Verlag, 2005 4 Seely R D, Goffredo M, Carter J N, Nixon M S. View invariant gait recognition. Handbook of Remote Biometrics for Surveillance and Security. London: Springer, 2009. 61 81 5 Perry J. Gait Analysis: Normal and Pathological Function. New Jersey: SLACK, 1992. 131 142 6 Vaughan C L, Davis B L, O Connor J C. Dynamics of Human Gait. Cape Town: Kiboho Publishers, 1999. 7 14 7 Inman V T, Ralston H J, Todd F. Human Walking. Baltimore: Williams and Wilkins, 1981 8 Huang P S, Harris C J, Nixon M S. Recognising humans by gait via parametric canonical space. Artificial Intelligence in Engineering, 1999, 13(4): 359 366 9 Boyd J E. Synchronization of oscillations for machine perception of gaits. Computer Vision and Image Understanding, 2004, 96(1): 35 59 10 Zhang R, Vogler C, Metaxas D. Human gait recognition at sagittal plane. Image and Vision Computing, 2007, 25(3): 321 330 11 Ben A C, Cutler R, Davis L. View-invariant estimation of height and stride for gait recognition. In: Proceedings of the International ECCV 2002 Workshop Copenhagen on Biometric Authentication. London, UK: Springer-Verlag, 2002. 155 167
36 37 12 Jia Li-Hao, Zou Jian-Hua. Improved background subtraction based on novel thresholding approach. Acta Automatica Sinica, 2009, 35(4): 394 400 (,.., 2009, 35(4): 394 400) 13 Shi J B, Tomasi C. Good features to track. In: Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition. Seattle, USA: IEEE, 1994. 593 600 14 Bradski G, Kaehler A. Learning OpenCV: Computer Vision with the OpenCV Library. Sebastopol: O Reilly Media, 2008. 214 219 15 Lucas B D, Kanade T. An iterative image registration technique with an application to stereo vision. In: Proceedings of the 7th International Joint Conference on Artificial Intelligence. Vancouver, Canada: Morgan Kaufmann Publishers, 1981. 674 679 16 Matthews I, Ishikawa T, Baker T. The template update problem. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2004, 26(6): 810 815 17 Haritaoglu I, Harwood D, Davis L S. W4: real-time surveillance of people and their activities. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2000, 22(8): 809 830 18 Leedan Y, Meer P. Heteroscedastic regression in computer vision: problems with bilinear constraint. International Journal of Computer Vision, 2000, 37(2): 127 150 19 Isard M, Blake A. Condensation-conditional density propagation for visual tracking. International Journal of Computer Vision, 1998, 29(1): 5 28 20 Georgescu B, Meer P. Point matching under large image deformations and illumination changes. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2004, 26(6): 674 688 21 Chang C C, Lin C J. Libsvm: a library for support vector machines [Online], available: http://www.csie.ntu.edu.tw/ cjlin/libsvm/, November 1, 2009... E-mail: lh.jia@stu.xjtu.edu.cn (JIA Li-Hao Ph. D. candidate at the Systems Engineering Institute, Xi an Jiaotong University. His research interest covers computer vision, pattern recognition, and video surveillance. Corresponding author of this paper.).. E-mail: jhzou@sei.xjtu.edu.cn (ZOU Jian-Hua Professor at the Systems Engineering Institute, Xi an Jiaotong University. His research interest covers image processing, complex systems analysis, and networked control systems.).. E-mail: chekai1985@gmail.com (CHE Kai Master student at the Systems Engineering Institute, Xi an Jiaotong University. His research interest covers computer vision, pattern recognition, and video surveillance.)