今年來為了安全的緣故各場合都開始架設起監視系統以備不時之需，但到意外真的發生的時候卻常使用人力來辨別不同攝影機間的同一位人物，我們期望改善這項缺失，由系統幫我們追蹤不同攝影機裡的行人。本篇論文提出了以條件機隨機場模型將跨攝影機追蹤視為尋找最小代價的問題。我們使用行人的外觀與行人移動耗費的時間為線索，結合群眾心理學推測行人的移動行為後定義限制建立出條件隨機場圖形，在外觀上使用雙胞胎卷積神經網路進行辨別，在行人移動耗費時間上以常態分布建模，將兩個模型的結果分別視為點與邊的代價。藉由前面所定義的限制運行我們提出的演算法，求出擁有最小耗費的配對。然後我們使用應用於多篇論文的資料集與評量方法比較此篇論文提出的方法與其他方法的效果，並確立同時使用行人的外貌與移動時間訊息比單只看外貌的效果更好。

In order to ensure safety, most public spaces now deploy monitor systems from time to time. However, in most scenarios the tracking works are executed manually when it is necessary. We expect this work to be done automatically by the system. This research proposes a method using conditional random field to formulate this problem as cost minimization problem. We take the appearance of pedestrians and their consumed time crossing camera as clues. Combine with crowd psychology to define constrains, constructing conditional random field graph. This research uses Siamese Convolutional Neural Network to recognize pedestrians’ appearance. Model the spending time pedestrian consume across camera by normal distribution. Take results of two models as the cost of nodes and edges. Apply the algorithm we proposed under constrains, finding matches with minimum cost. We compare the accuracy of this research with other methods using a common datasets and benchmark. Confirm the effect of using appearance and spatio-temporal clues is better than using the appearance alone.

[1] A. F. Aveni, "The not-so-lonely crowd: Friendship groups in collective behavior," Sociometry, 1977, pp.96-99
[2] X. Chen and B. Bhanu, "Integrating social grouping for multitarget tracking across cameras in a CRF model," IEEE Transactions on Circuits and Systems for Video Technology, 27(11), 2017, pp. 2382-2394
[3] W. Chen, L. Cao, X. Chen and K. Huang, "An equalized global graph model-based approach for multicamera object tracking." IEEE Transactions on Circuits and Systems for Video Technology, 27(11), 2017, pp.2367-2381
[4] X. Chen, A. Le and B. Bhanu, "Multitarget tracking in nonoverlapping cameras using a reference set," IEEE Sensors Journal, 15(5), 2015, pp.2692-2704
[5] X. Chen, K. Huang and T. Tan, "Object tracking across non-overlapping views by learning inter-camera transfer models," Pattern Recognition, 47(3), 2014, pp.1126-1137
[6] S. Chopra, R. Hadsell and Y. LeCun, "Learning a similarity metric discriminatively, with application to face verification," IEEE Computer Society Conference on Computer Vision and Pattern Recognition 2005(CVPR 2005), Vol. 1, 2005
[7] W. Ge, T. R. Collins and R. B. Ruback, "Vision-based analysis of small groups in pedestrian crowds," IEEE Tansactions on Pattern Analysis and Machine Intelligence, 34(5), 2012, pp.1003-1016
[8] R. Hadsell, S. Chopra, and Y. LeCun, "Dimensionality reduction by learning an invariant mapping," 2006 IEEE Computer Society Conference on Computer Vision and Pattern Recognition (CVPR2006), 2006, pp. 1735-1742
[9] A. Heili, C. Chen, and J.-M. Odobez, "Detection-based multi-human tracking using a CRF model," IEEE International Conference on Computer Vision Workshops (ICCV Workshops), 2011
[10] D. Helbing and P. Molnar, "Social force model for pedestrian dynamics," Physical review E, 51(5), 1995, pp. 4282
[11] O. Javed, Z. Rasheed, K. Shafique and M. Shah, “Tracking across multiple cameras with disjoint views,” Proceedings of the Ninth IEEE International Conference on Computer Vision, 2008, pp. 952
[12] V. Kolmogorov and R. Zabih, "What Energy Functions Can Be Minimized via Graph Cut?", IEEE Transactions on Patten Analysis and Machine Intelligence, Vol 2, 2004
[13] D. Makris, T. Ellis and J. Black, "Bridging the gaps between cameras," Proceedings of the 2004 IEEE Computer Society Conference on Computer Vision and Pattern Recognition(CVPR 2004), Vol. 2, 2004
[14] M. Moussaïd , N. Perozo, S. Garnier, D. Helbing and G. Theraulaz, "The walking behaviour of pedestrian social groups and its impact on crowd dynamics," Public Library of Science(PLOS one), 5(4), 2010
[15] E. Reinhard, M. Adhikhmin and B. Gooch, "Color transfer between images." IEEE Computer Graphics and Applications, 21(5), 2001, pp.34-41
[16] D. L. Ruderman, T. W. Cronin and C.-C. Chiao, "Statistics of cone responses to natural images: implications for visual coding," Journal of the Optical Society of America A(JOSA A), 15(8) , 1998, pp.2036-2045
[17] H. B. Shitrit, J. Berclaz, F. Fleuret and P. Fua, "Multi-commodity network flow for tracking multiple people," IEEE Transactions on Pattern Analysis and Machine Intelligence, 36(8), 2014, pp.1614-1627
[18] Y. T. Tesfaye, E. Zemene, A. Prati, M. Pelillo and M. Shah, "Multi-target tracking in multiple non-overlapping cameras using constrained dominant sets," IEEE, 2017
[19] B. Yang and R. Nevatia, "An online learned CRF model for multi-target tracking." IEEE Conference on Computer Vision and Pattern Recognition (CVPR2012), 2012
[20] S. Zagoruyko and N. Komodakis, "Learning to compare image patches via convolutional neural networks," Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, 2015
[21] NLPR MCT Dataset. [Online]. Available: http://mct.idealtest.org/ Datasets.html