人物辨識與追蹤的技術常被應用於安全監控系統，此系統通常會同時監控多支攝影機的拍攝畫面。為了有效達到安全監控的目的，此系統必須在多支不同的攝影機之間準確的追蹤入侵人物。在本論文中，我們設計一個跨攝影機的人物追蹤 (Cross-Camera Human Tracking, CCHT) 演算法，此演算法會根據入侵人物於畫面中的比例來辨別攝影機的高低。為了進行跨攝影機的人物追蹤，我們在CCHT演算法中設計三個模組，第一個模組用來計算畫面的封包遺失率，第二個模組用來計算入侵人物於畫面中的比例，第三個模組用來擷取人物的身體特徵。根據畫面的封包遺失率與入侵人物於畫面中的比例 (此比例可以表示攝影機的高低)，我們分為三種不同的解決方法，第一種方法為使用身體特徵對入侵人物進行辨識與追蹤，第二種方法為同時使用人臉特徵與身體特徵對入侵人物進行辨識與追蹤，第三種方法為使用人臉特徵對入侵人物進行辨識與追蹤。最後為了驗證我們提出的CCHT演算法，我們設計信心分數的統計來分析與比較跨攝影機的人物追蹤準確率。另外，為了比較在5G與Wi-Fi網路中有無使用CCHT演算法的差異，我們在Mavenir的5G-SA-Open RAN與Wi-Fi的網路環境下實作CCHT演算法。

Human identification and tracking technology has been widely used in security monitoring system, which may employ multiple cameras at the same time. For effective security monitoring, the system must accurately identify the intruder and then track the intruder across multiple different cameras. In this thesis, we design a Cross-Camera Human Tracking (CCHT) algorithm, which first needs to determine whether the camera is set in high or low position based on the proportion of the intruder in the frame. For cross-camera human tracking, we design three modules in the CCHT algorithm; the first module is used to calculate the packet loss rate of a frame, the second module is used to calculate the proportion of the intruder in the frame, and the third module is used to extract the body features of the intruder. According to the packet loss rate of a frame and the proportion of the intruder in the frame (this proportion can represent whether the camera is set in high or low position), we can divide the problem and come out with three solutions; the first solution uses body features to track the intruders, the second solution uses both facial features and body features to track the intruders, and the third solution uses facial features to track the intruders. Finally, in order to validate the proposed CCHT algorithm, we design confidence-score statistics to analyze and compare human tracking accuracy across two cameras. In addition, in order to compare the tracking accuracy between 5G and Wi-Fi networks with and without using the CCHT algorithm, we implement the CCHT algorithm in Mavenir's 5G-SA-Open RAN and Wi-Fi networks.

論文審定書 i
誌謝 ii
摘要 iii
Abstract iv
目錄 v
圖目錄 vii
表目錄 viii
第一章 導論 1
1.1 研究動機 1
1.2 研究方法 1
1.3 章節介紹 3
第二章 使用影音串流的人物追蹤 4
2.1 影音串流的解壓縮 4
2.2 YOLO軟體 7
2.2.1 YOLO的特徵訓練 7
2.2.2 YOLO的辨識流程 8
2.2.3 YOLO的信心分數 9
2.3 相關研究 10
第三章 跨多支攝影機的人物追蹤 14
3.1 跨多支攝影機追蹤的網路架構 14
3.2 跨攝影機的人物追蹤演算法 15
3.2.1 影音串流中I畫面的遺失率 (IFLR) 18
3.2.2 入侵人物於畫面中的比例 (HFR) 20
3.2.3 身體特徵的計算 21
3.2.3.1 色彩的平均 21
3.2.3.2 人物的長寬比 23
3.2.3.3 手部與腳部的最大擺動角度 23
3.2.4 修改的信心分數 25
第四章 實作與結果分析 27
4.1 實驗拓撲與設備規格 27
4.2 在YOLO Server上的實作 31
4.2.1 人物的訓練 31
4.2.2 IFLR計算模組的虛擬碼 33
4.2.3 HFR計算模組的虛擬碼 36
4.2.4 BFC模組的虛擬碼 38
4.2.5 信心分數的統計 44
4.3 實作結果與分析 47
4.3.1 實驗情境與參數設定 47
4.3.2 結果分析 48
第五章 結論與未來工作 51
5.1 結論 51
5.2 遭遇問題 52
5.3 未來工作 52
Reference 53
Acronyms 58
Index 60

[1] Rob Koenen, "MPEG-4 Overview-(V.21–Jeju Version)," ISO/IEC JTC1/SC29/WG11 N4668, Mar. 2002.
[2] T. Sikora, "Trends and Perspectives in Image and Video Coding," Proceedings of the IEEE, Vol. 93, No. 1, pp. 6-17, Jan. 2005.
[3] D. A. Huffman, "A Method for the Construction of Minimum-Redundancy Codes," Proceedings of the IRE, Vol. 40, No. 9, pp. 1098-1101, Sep. 1952.
[4] K. L. Chung and J. G. Wu, "Level-Compressed Huffman Decoding," IEEE Transactions on Communications, Vol. 47, No. 10, pp. 1455-1457, Oct. 1999.
[5] H. Yang, H. Chen, J. Chen, S. Esenlink, S. Sethuraman, X. Xiu, E. Alshina, and J. Luo, "Subblock-Based Motion Derivation and Inter Prediction Refinement in the Versatile Video Coding Standard," IEEE Transactions on Circuits and Systems for Video Technology, Vol. 31, No. 10, pp. 3862-3877, Oct. 2021.
[6] G. Choi, P. G. Heo, and H. Park, "Triple-Frame-Based Bi-Directional Motion Estimation for Motion-Compensated Frame Interpolation," IEEE Transactions on Circuits and Systems for Video Technology, Vol. 29, No. 5, pp. 1251-1258, May 2019.
[7] Y. H. Chen, C. C. Cheng, T. D. Chuang, C. Y. Chen, S. Y. Chien, and L. G. Chen, "Efficient Architecture Design of Motion-Compensated Temporal Filtering/Motion Compensated Prediction Engine," IEEE Transactions on Circuits and Systems for Video Technology, Vol. 18, No. 1, pp. 98-109, Jan. 2008.
[8] L. Jiao, F. Zhang, F. Liu, S. Yang, L. Li, Z. Feng, and R. Qu, "A Survey of Deep Learning-Based Object Detection," IEEE Access, Vol. 7, pp. 128837-128868, Sep. 2019.
[9] Z. -Q. Zhao, P. Zheng, S. -T. Xu, and X. Wu, "Object Detection with Deep Learning: A Review," IEEE Transactions on Neural Networks and Learning Systems, Vol. 30, No. 11, pp. 3212-3232, Nov. 2019.
[10] E. Bochinski, V. Eiselein, and T. Sikora, "High-Speed Tracking-by-Detection without Using Image Information," 2017 14th IEEE International Conference on Advanced Video and Signal Based Surveillance (AVSS), Lecce, Italy, pp. 1-6, 29 Aug.-1 Sep. 2017.
[11] N. Wojke, A. Bewley, and D. Paulus, "Simple Online and Realtime Tracking with a Deep Association Metric," arXiv preprint arXiv:1703.07402, Mar. 2017.
[12] Z. Kalal, K. Mikolajczyk, and J. Matas, "Tracking-Learning-Detection," IEEE Transactions on Pattern Analysis and Machine Intelligence, Vol. 34, No. 7, pp. 1409-1422, Jul. 2012.
[13] A. Bewley, Z. Ge, L. Ott, F. Ramos, and B. Upcroft, "Simple Online and Realtime Tracking," 2016 IEEE International Conference on Image Processing (ICIP), Phoenix, AZ, USA, pp. 3464-3468, 25-28 Sep. 2016.
[14] J. Redmon, S. Divvala, R. Girshick, and A. Farhadi, "You Only Look Once: Unified, Real-Time Object Detection," 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR), Las Vegas, NV, USA, pp. 779-788, 27-30 Jun. 2016.
[15] J. Redmon and A. Farhadi, "YOLO9000: Better, Faster, Stronger," Proceedings of the IEEE conference on computer vision and pattern recognition, Honolulu, pp. 7263-7271, 21-26 Jul. 2017.
[16] J. Redmon and A. Farhadi. "YOLOv3: An Incremental Improvement," arXiv preprint arXiv:1804.02767, Apr. 2018.
[17] A. Bochkovskiy, C. -Y. Wang, and H. -Y. Liao, "YOLOv4: Optimal Speed and Accuracy of Object Detection," arXiv preprint arXiv:2004.10934, Apr. 2020.
[18] R. Girshick, J. Donahue, T. Darrell, and J. Malik, "Rich Feature Hierarchies for Accurate Object Detection and Semantic Segmentation," Proceedings of the IEEE conference on computer vision and pattern recognition, Columbus, OH, USA, pp. 580-587, 23-28 Jun. 2014.
[19] J. R. Uijlings, K. E. Sande, T. Gevers, and A. W. Smeulders, "Selective Search for Object Recognition," International Journal of Computer Vision, Vol. 104, No. 2, pp. 154-171, Sep. 2013.
[20] T. Y. Lin, P. Dollár, R. Girshick, K. He, B. Hariharan, and S. Belongie, "Feature Pyramid Networks for Object Detection," Proceedings of the IEEE conference on computer vision and pattern recognition, Honolulu, HI, USA, pp. 2117-2125, 21-26 Jul. 2017.
[21] S. K. Teoh, V. V. Yap, and H. Nisar, "A Non-Overlapping View Human Tracking Algorithm Using HSV Colour Space," 2019 International Conference on Green and Human Information Technology (ICGHIT), Kuala Lumpur, Malaysia, pp. 97-102, 15-17 Jan. 2019.
[22] J. Kwak, B. C. Ko, and J. Y. Nam, "Pedestrian Tracking Using Online Boosted Random Ferns Learning in Far-Infrared Imagery for Safe Driving at Night," IEEE Transactions on Intelligent Transportation Systems, Vol. 18, No. 1, pp. 69-81, Jan. 2017.
[23] M. Morimoto and M. Yukitou, "A Recognition Method for Overlapped Objects Using Multiple RGB-D Sensors," 2018 World Automation Congress (WAC), Stevenson, WA, USA, pp. 1-5, 3-6 Jun. 2018.
[24] X. Chen, L. An, and B. Bhanu, "Reference Set Based Appearance Model for Tracking Across Non-Overlapping Cameras," 2013 Seventh International Conference on Distributed Smart Cameras (ICDSC), Palm Springs, CA, pp. 1-6, 29 Oct.-1 Nov. 2013.
[25] C. Chu, J. Hwang, K. Lan, and S. Wang, "Tracking Across Multiple Cameras with Overlapping Views Based on Brightness and Tangent Transfer Functions," 2011 Fifth ACM/IEEE International Conference on Distributed Smart Cameras, Ghent, Belgium, pp. 1-6, 22-25 Aug. 2011.
[26] S. You, H. Yao, and C. Xu, "Multi-Target Multi-Camera Tracking with Optical-Based Pose Association," IEEE Transactions on Circuits and Systems for Video Technology, Vol. 31, No. 8, pp. 3105-3117, Aug. 2021.
[27] Y. J. Cho and K. J. Yoon, "Distance-Based Camera Network Topology Inference for Person Re-identification," Pattern Recognition Letters, Vol. 125, pp. 220-227, Jul. 2019.
[28] M. Yaghi, T. Basmaji, R. Salim, J. Yousaf, H. Zia, and M. Ghazal, "Real-Time Contact Tracing During a Pandemic Using Multi-camera Video Object Tracking," 2020 International Conference on Decision Aid Sciences and Application (DASA), Sakheer, Bahrain, pp. 872-876, 8-9 Nov. 2020.
[29] S. Jain, X. Zhang, Y. Zhou, G. Ananthanarayanan, J. Jiang, Y. Shu, P. Bahl, and J. Gonzalez, "Spatula: Efficient Cross-Camera Video Analytics on Large Camera Networks," 2020 IEEE/ACM Symposium on Edge Computing (SEC), San Jose, CA, USA, pp. 110-124, 12-14 Nov. 2020.
[30] Y. He, X. Wei, X. Hong, W. Shi and Y. Gong, "Multi-Target Multi-Camera Tracking by Tracklet-to-Target Assignment," IEEE Transactions on Image Processing, Vol. 29, pp. 5191-5205, Mar. 2020.