物件追蹤的技術近年來被廣泛應用在治安的監控，而目前最廣泛使用的物件追蹤軟體是YOLOv4，但是當入侵物件離開地面固定式攝影機的監控範圍後，就無法繼續被追蹤，因此，無人機被使用來繼續追蹤入侵物件。目前的無人機都是需要先輸入全球衛星定位系統(Global Positioning System, GPS)的座標才能抵達入侵物件的位置，但是入侵物件不會主動提供GPS座標給無人機，所以本論文提出一個GPS的估算與動態切換攝影機(Dynamic Switching Cameras, DSC)的物件追蹤機制，GPS的估算包含地面固定式攝影機與無人機移動式攝影機的GPS估算模組，在地面固定式攝影機的GPS估算模組中，我們會根據參考點的實際位置、入侵物件與參考點在地面固定式攝影機畫面中的像素差值來估算入侵物件的近似GPS座標。無人機移動式攝影機的GPS估算模組可以分為三個部分，第一個部分是無人機移動式攝影機與地面固定式攝影機畫面重疊之參考點的實際位置，第二個部分是兩支攝影機畫面解析度的比例，第三個部分是入侵物件在無人機移動式攝影機畫面中的相對座標。另外，攝影機切換的時機須滿足兩個條件，第一個條件是當無人機已抵達入侵物件的近似GPS座標時，第二個條件是當入侵物件即將離開地面固定式攝影機的監控範圍時。最後，我們在Mavenir 5G-SA-Open RAN平台實作本論文的DSC機制，我們在YOLO伺服器(YOLO Server)設計兩種攝影機的GPS估算模組，並增加切換攝影機的判斷條件。在實作完成後，我們會量測並分析GPS估算模組的處理時間，在不同的行動通訊網路中，我們也比較地面固定式攝影機與無人機移動式攝影機畫面的切換時間。

Object tracking technology has been widely used in security monitoring in recent years. One of the most widely used object tracking software is YOLOv4. However, when the trespassing object leaves the monitoring range of an IP camera on the ground, the object may not be continuously tracked. Therefore, one may use a drone to continue to track the trespassing object in the air. Currently, a drone requires the assistance of GPS (Global Positioning System) to reach the location of the trespassing object. The problem is that a trespassing object may not actively provide its GPS to a drone. Therefore, in this thesis, we propose a GPS estimation and dynamic switching cameras (DSC) mechanism for the object tracking. We have designed two GPS estimation modules, one for ground IP cameras, and the other one for drone mobile cameras (DMC). The former module can estimate the trespassing object's approximate GPS (OAGPS) based on the GPS of three reference points on the ground. The latter module is divided into three parts; the first part is the GPS of the reference points where the images of DMC and the ground IP camera overlaps, the second part is the ratio of two cameras' image resolution, and the third part is the object's pixel position in the DMC frame. Additionally, the time to switch the two cameras must meet two conditions; the first condition is when the drone has reached the OAGPS location, and the second condition is when the trespassing object is about to leave the monitoring range of the ground IP camera. Finally, we implement the DSC mechanism on a Mavenir 5G-SA-Open RAN platform. After the implementations, we measure and analyze the processing time of the two GPS estimation modules. We also compare the picture switch-over time between the fixed and the mobile cameras (PSOT) in different mobile communication networks.

論文審定書…………………………………………………………………………….i
致謝…………………………………………………………………….......………….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 物件的追蹤……………………………………………….………………….4
2.1 影像的物件追蹤…………………………………………….…………………4
2.1.1 YOLO的Detector ………………………………………….……………...5
2.1.2 YOLO的Tracker………………………………………….………………..6
2.2 GPS的物件追蹤…………………………………………….………………….9
2.3 相關研究…………………………………………….………………………..11
第三章 GPS的估算與動態切換攝影機……………………………………………15
3.1 DSC機制的系統架構…………………………………………….…………...15
3.2 GPS的估算模組…………………………………………….………………...18
3.2.1 GEM-FC…………………………………………….…………………….18
3.2.2 GEM-MC…………………………………………….……………………23
3.3 DSC機制的運作流程…………………………………………….…………...26
3.3.1 在YOLO Server的運作流程…………………………………………….26
3.3.2 在樹莓派的運作流程…………………………………………….……...27
第四章 實作與結果分析…………………………………………….……………...29
4.1 在YOLO Server的實作…………………………………………….………...29
4.1.1 GEM-FC的虛擬碼…………………………………………….………….29
4.1.2 GEM-MC的虛擬碼…………………………………………….………...34
4.2實驗環境與設備規格…………………………………………….…………...36
4.3實作的結果分析…………………………………………….…………….......40
4.3.1 GEM-FC與GEM-MC的處理時間………………………………………40
4.3.2 PSOT…………………………………………….………………………..41
4.3.2.1 ITMC…………………………………………….…………………...42
4.3.2.2 NDDS與NDSY…………………………………………….………..46
4.3.2.3 QDSS與QDYS…………………………………………….………...46
4.3.2.4 量測結果…………………………………………….………………48
第五章 結論與未來工作…………………………………………….……………...50
5.1 結論………………………………………….………………………………..50
5.2 遭遇困難………………………………………….…………………………..51
5.3 未來工作…………………………………………….………………………..52
References…………………………………………….……………………………...53
Acronyms…………………………………………….………………………………57
Index…………………………………………….……………………………………59

[1] 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.
[2] J. Redmon and A. Farhadi. "YOLOv3: An incremental improvement," arXiv preprint arXiv:1804.02767, Apr. 2018.
[3] A. Bochkovskiy, C. -Y. Wang, and H. -Y. Liao, "YOLOv4: Optimal Speed and Accuracy of Object Detection," arXiv preprint arXiv:2004.10934, Apr. 2020.
[4] P. Enge and P. Misra, "Special Issue on Global Positioning System," Proceedings of the IEEE, vol. 87, no. 1, pp. 3-15, Jan. 1999.
[5] Wei Liu, Dragomir Anguelov, Dumitru Erhan, Christian Szegedy, Scott Reed, Cheng-Yang Fu, and Alexander C. Berg, "SSD: Single Shot MultiBox Detector," arXiv preprint arXiv:1512.02325, Dec. 2016.
[6] G. Singh, S. Saha, M. Sapienza, P. Torr, and F. Cuzzolin, "Online Real-Time Multiple Spatiotemporal Action Localisation and Prediction," 2017 IEEE International Conference on Computer Vision (ICCV), Venice, Italy, pp. 3657-3666, 22-29 Oct., 2017.
[7] K. Duan, S. Bai, L. Xie, H. Qi, Q. Huang, and Q. Tian, "CenterNet: Keypoint Triplets for Object Detection," 2019 IEEE/CVF International Conference on Computer Vision (ICCV), Seoul, Korea, pp. 6568-6577, 27 Oct.-2 Nov., 2019.
[8] T. -Y. Lin, P. Goyal, R. Girshick, K. He, and P. Dollár, "Focal Loss for Dense Object Detection," 2017 IEEE International Conference on Computer Vision (ICCV), Venice, Italy, pp. 2999-3007, 22-29 Oct., 2017.
[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] 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.
[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] 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.
[13] S. Mahfouz, F. Mourad-Chehade, P. Honeine, J. Farah, and H. Snoussi, "Target Tracking Using Machine Learning and Kalman Filter in Wireless Sensor Networks," IEEE Sensors Journal, vol. 14, no. 10, pp. 3715-3725, Oct. 2014.
[14] H. -W. Kuhn, "The Hungarian method for the assignment problem," Naval Research Logistics Quarterly, Vol. 2, 1955, pp. 83-97., Mar. 1955.
[15] T. Weiss, N. Kaempchen, and K. Dietmayer, "Precise ego-localization in urban areas using Laserscanner and high accuracy feature maps," IEEE Proceedings. Intelligent Vehicles Symposium, Las Vegas, NV, USA, pp. 284-289, 6-8 Jun., 2005.
[16] 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.
[17] S. Kim and J. Choi, "Face Recognition Method Based on Fixed and PTZ Camera Control for Moving Humans," 2019 Eleventh International Conference on Ubiquitous and Future Networks (ICUFN), Zagreb, Croatia, pp. 561-563, 2-5 Jul., 2019.
[18] J. Park, D. H. Kim, Y. S. Shin, and S. Lee, "A comparison of convolutional object detectors for real-time drone tracking using a PTZ camera," 2017 17th International Conference on Control, Automation and Systems (ICCAS), Jeju, Korea (South), pp. 696-699, 18-21 Oct., 2017.
[19] S. Atoev, A. Nishanov, and F. Abdirazakov, "Object Tracking Method Based on Kalman Filter and Camshift Algorithm for UAV Applications," 2021 International Conference on Information Science and Communications Technologies (ICISCT), Tashkent, Uzbekistan, pp. 1-4, 3-5 Nov., 2021.
[20] D. Xing, N. Evangeliou, A. Tsoukalas, and A. Tzes, "Siamese Transformer Pyramid Networks for Real-Time UAV Tracking," 2022 IEEE/CVF Winter Conference on Applications of Computer Vision (WACV), Waikoloa, HI, USA, pp. 1898-1907, 3-8 Jan., 2022.
[21] H. A. Abdallah Dafallah, "Design and implementation of an accurate real time GPS tracking system," The Third International Conference on e-Technologies and Networks for Development (ICeND2014), 2014, pp. 183-188, Beirut, Lebanon, 29 Apr.-1 May, 2014.
[22] F. N. Ameen, Z. S. Mohammed, and A. I. Sddiq, "An Economic Tracking Scheme for GPS-GSM Based Moving Object Tracking System," 2018 2nd International Conference for Engineering, Technology and Sciences of Al-Kitab (ICETS), Karkuk, Iraq, pp. 28-32, 4-6 Dec., 2018.
[23] Wahyudi, M. S. Listiyana, Sudjadi, and Ngatelan, "Tracking Object based on GPS and IMU Sensor," 2018 5th International Conference on Information Technology, Computer, and Electrical Engineering (ICITACEE), Semarang, Indonesia, pp. 214-218, 27-28 Sep., 2018.
[24] K. L. K. Sudheera, G. G. N. Sandamali, W. N. D. C. Sandaruwan, and N. D. Jayasundere, "Offline tracking system for deep sea going vessels using GPS and GPRS," 2015 8th International Conference on Ubi-Media Computing (UMEDIA), Colombo, Sri Lanka, pp. 55-60, 24-26 Aug., 2015.
[25] J. Lee, S. Moon, D. -W. Nam, J. Lee, A. R. Oh, and W. Yoo, "A Study on American Football Player Tracking Based on Video Through Deep Learning and GPS Convergence," 2021 International Conference on Information and Communication Technology Convergence (ICTC), Jeju Island, Korea, pp. 1114-1116, 20-22 Oct., 2021.