在本篇論文中，我們分別探討重疊與非重疊多攝影機影像之人物追蹤。針對非重疊攝影機之人物追蹤，我們使用目前最大且最完整的資料集 DukeMTMCT，並深入研究在此資料集擁有良好效果之方法 DeepCC。我們發現到 DeepCC 所使用的偵測器 OpenPose 容易有錯誤的偵測產生。根據這個問題，我們提出一個背景相似度比對演算法，來計算偵測框與原始背景間的相似度，進而移除 OpenPose 所產生的錯誤偵測結果。此外，我們還觀察到明顯不同的人物有被指定為同一個 ID 的現象。針對此問題，我們設計了一種軌跡修復的策略來校正這種因為不穩定聚類方法所產生的錯誤聚類現象。我們所提出的方法在國際上具權威的多物件追蹤挑戰賽中具有優越的成果。
針對重疊攝影機之人物追蹤，我們在球場蒐集了籃球員跑動之重疊攝影 機資料集。我們針對該資料集設計了一套穩定追蹤球員的演算法。不同於非重疊攝影機，在每一個攝影機分別偵測完人物後，需要使用偵測體聚類演算法將多台攝影機之偵測結果投影到相同的座標系統，並根據座標資訊融合同 一人物之多偵測體。我們提出一個可以準確將同一人物之偵測體整合之偵測體聚類架構。此外，由於籃球員的移動較行人快速且難以預測，我們使用區域一致性假定來衡量球員移動上的關聯性，並使用一個一對一聚類演算法來產生球員軌跡。我們提出的偵測演算法可以很準確地偵測籃球員的跑動軌跡。

In this work, we explore the challenges of multiple human tracking in overlap­ ping and non­overlapping multiple cameras. For non­overlapping multi­camera hu­ man tracking, we use DukeMTMCT which is the largest and most completely la­ beled dataset in Multi­Target Multi­Camera Tracking (MTMCT). We investigate a state­of­the­art work on DukeMTMCT named DeepCC, and dig out two main problems. The first problem is that the OpenPose is prone to false detection, which seriously affects performance. The second problem is that two different persons may be assigned with the same ID. According to the corresponding problems, we not only propose a method to measure the similarity between detected bounding box and its original background avoiding false detection caused by OpenPose, but also design a strategy to correct the tracking trajectories which are affected by the unreliability of the correlation matrix clustering method proposed by DeepCC. Our method outperforms the state­of­the­art on DukeMTMCT.
For overlapping multi­camera human tracking, we propose a robust basketball player tracking framework for multi­cameras which have high portion of overlap­ ping with each other and are set at human height. A novel detection grouping method is proposed to more correctly merge the projected detection results. Instead of us­ ing linear motion assumption to predict the human motion, we applied a regional consistency assumption to calculate the motion affinity. Furthermore, we design a one­to­one clustering method to associate the most matching tracklets together and generate final trajectory results. Since there is no public labeled overlapping multi­cameras basketball dataset, we collected our own dataset, MISBasketball, and labeled the ground truth to evaluate the proposed tracking framework.

[1] Sportuv. https://www.stats.com/sportvu­basketball/.
[2] J. Berclaz, F. Fleuret, E. Turetken, and P. Fua. Multiple object tracking us­ ing k­shortest paths optimization. IEEE transactions on pattern analysis and machine intelligence, 33(9):1806–1819, 2011.
[3] K. Bernardin and R. Stiefelhagen. Evaluating multiple object tracking per­ formance: the clear mot metrics. Journal on Image and Video Processing, 2008:1, 2008.
[4] Z. Cao, T. Simon, S.­E. Wei, and Y. Sheikh. Realtime multi­person 2d pose estimation using part affinity fields. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, pages 7291–7299, 2017.
[5] H.­S. Fang, S. Xie, Y.­W. Tai, and C. Lu. RMPE: Regional multi­person pose estimation. In ICCV, 2017.
[6] F. Fleuret, J. Berclaz, R. Lengagne, and P. Fua. Multicamera people tracking with a probabilistic occupancy map. IEEE transactions on pattern analysis and machine intelligence, 30(2):267–282, 2008.
[7] R. Henschel, L. Leal­Taixé, D. Cremers, and B. Rosenhahn. Fusion of head and full­body detectors for multi­object tracking. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition Workshops, pages 1428–1437, 2018.
[8] M.­C. Hu, M.­H. Chang, J.­L. Wu, and L. Chi. Robust camera calibration and player tracking in broadcast basketball video. IEEE Transactions on Multime­ dia, 13(2):266–279, 2010.
[9] N. Jiang, S. Bai, Y. Xu, C. Xing, Z. Zhou, and W. Wu. Online inter­camera trajectory association exploiting person re­identification and camera topology.In 2018 ACM Multimedia Conference on Multimedia Conference, pages 1457– 1465. ACM, 2018.
[10] S. C. Johnson. Hierarchical clustering schemes. Psychometrika, 32(3):241– 254, 1967.
[11] L. Leal­Taixé, A. Milan, I. Reid, S. Roth, and K. Schindler. Motchallenge 2015: Towards a benchmark for multi­target tracking. arXiv preprint arXiv: 1504.01942, 2015.
[12] S. Lloyd. Least squares quantization in pcm. IEEE transactions on informa­ tion theory, 28(2):129–137, 1982.
[13] W.­L. Lu, J.­A. Ting, J. J. Little, and K. P. Murphy. Learning to track and identify players from broadcast sports videos. IEEE transactions on pattern analysis and machine intelligence, 35(7):1704–1716, 2013.
[14] A. Maksai, X. Wang, F. Fleuret, and P. Fua. Non­markovian globally consis­ tent multi­object tracking. In Proceedings of the IEEE International Confer­ ence on Computer Vision, pages 2544–2554, 2017.
[15] A. Y. Ng, M. I. Jordan, and Y. Weiss. On spectral clustering: Analysis and an algorithm. In Advances in neural information processing systems, pages 849–856, 2002.
[16] S. Ren, K. He, R. Girshick, and J. Sun. Faster r­cnn: Towards real­time object detection with region proposal networks. In Advances in neural information processing systems, pages 91–99, 2015.
[17] E. Ristani, F. Solera, R. Zou, R. Cucchiara, and C. Tomasi. Performance mea­ sures and a data set for multi­target, multi­camera tracking. In European Con­ ference on Computer Vision, pages 17–35. Springer, 2016.
[18] E. Ristani and C. Tomasi. Features for multi­target multi­camera tracking and re­identification. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, pages 6036–6046, 2018.
[19] S. Tang, M. Andriluka, B. Andres, and B. Schiele. Multiple people tracking by lifted multicut and person re­identification. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, pages 3539–3548, 2017.
[20] Y. T. Tesfaye, E. Zemene, A. Prati, M. Pelillo, and M. Shah. Multi­target tracking in multiple non­overlapping cameras using constrained dominant sets. arXiv preprint arXiv:1706.06196, 2017.
[21] G. Wang, Y. Wang, H. Zhang, R. Gu, and J.­N. Hwang. Exploit the connectiv­ ity: Multi­object tracking with trackletnet. arXiv preprint arXiv:1811.07258, 2018.
[22] N. Wojke, A. Bewley, and D. Paulus. Simple online and realtime tracking with a deep association metric. In 2017 IEEE International Conference on Image Processing (ICIP), pages 3645–3649. IEEE, 2017.
[23] J. Xing, H. Ai, L. Liu, and S. Lao. Multiple player tracking in sports video: A dual­mode two­way bayesian inference approach with progressive observation modeling. IEEE Transactions on Image Processing, 20(6):1652–1667, 2010.
[24] K. Yoon, Y.­m. Song, and M. Jeon. Multiple hypothesis tracking algorithm for multi­target multi­camera tracking with disjoint views. IET Image Processing, 12(7):1175–1184, 2018.
[25] S.­I. Yu, D. Meng, W. Zuo, and A. Hauptmann. The solution path algorithm for identity­aware multi­object tracking. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, pages 3871–3879, 2016.
[26] Z. Zhang, J. Wu, X. Zhang, and C. Zhang. Multi­target, multi­camera track­ ing by hierarchical clustering: recent progress on dukemtmc project. arXiv preprint arXiv:1712.09531, 2017.