成功大學博碩士論文系統

簡易檢索 / 詳目顯示

回結果列表

研究生：	鮑子婕 Bao, Tzu-Chieh
論文名稱：	基於移動資訊與共享平面轉換之多相機多人追蹤研究 Motion Information and Shared Plane Transformation for Multi-Camera Multi-Person Tracking
指導教授：	呂學展 Lu, Hsueh-Chan
學位類別：	碩士 Master
系所名稱：	工學院 - 測量及空間資訊學系 Department of Geomatics
論文出版年：	2026
畢業學年度：	114
語文別：	英文
論文頁數：	108
中文關鍵詞：	多相機多物件追蹤、電腦視覺、單應性矩陣、物件偵測、軌跡關聯
外文關鍵詞：	Multi-camera multi-object tracking, computer vision, homography matrix, object detection, trajectory association
相關次數：	點閱：4 下載：0
分享至:	分享至facebook 分享至twitter

查詢本校圖書館目錄查詢臺灣博碩士論文知識加值系統勘誤回報

多相機多目標追蹤（Multi-Camera Multi-Object Tracking, MC-MOT）是一項透過多台相機協同運作，對場景中的目標進行持續追蹤與身份維護的電腦視覺技術，廣泛應用於智慧監控、交通管理及行為分析等領域。在智慧城市與人工智慧技術快速發展的趨勢下，多相機多目標追蹤已成為建構大規模智慧監控系統的重要基礎。其優勢包括能夠擴大監控範圍、降低單一相機視野受限的影響，以及提供目標於不同區域間的完整移動軌跡。然而系統運作過程中仍存在數項挑戰：不同相機之間通常具有視角差異、光照變化及解析度不一致等問題，使得同一目標在不同相機下呈現出明顯的外觀差異，增加跨相機身份關聯的困難；其次，當目標受到遮擋、短暫離開視野或與其他目標發生交互時，容易造成追蹤中斷或身份轉換，進而降低系統整體追蹤效能。為解決上述問題並提升多相機目標追蹤的穩定性與準確性，本研究將問題聚焦於三個關鍵面向：（1）共享平面上的空間幾何映射、（2）基於移動方向與距離的跨相機軌跡配對，以及（3）結合歷史身份資訊的跨時間身份一致性維護。針對多相機場景中的身份關聯困難問題，本研究提出一套兩階段跨相機配對方法，利用共享平面上的幾何資訊與目標移動特徵建立初步配對，並透過歷史身份資訊與群組一致性策略進行身份衝突處理與全域身份統一。實驗部分，本研究使用虛擬環境之多相機行人追蹤資料集進行評估，並評估各項指標以分析追蹤效能。在研究案例中我們的方法能通過移動資訊維持較穩定的關聯效果，甚至在某些情況下優於部分既有方法。研究成果證明，結合共享平面之幾何資訊與移動方向特徵在維持較低計算成本的同時，仍能獲得穩定且具競爭力的追蹤效能，不依賴高維度外觀特徵即可建立目標關聯，提升全域身分配對準確性與整體追蹤品質。

Multi-Camera Multi-Object Tracking (MC-MOT) is a computer vision technology that enables multiple cameras to collaboratively perform continuous target tracking and identity preservation across a monitored scene. It has been widely applied in intelligent surveillance, traffic management, and behavior analysis. With the rapid advancement of smart city initiatives and artificial intelligence technologies, MC-MOT has become a fundamental component of large-scale intelligent surveillance systems. Its advantages include expanding the monitoring coverage, reducing the limitations caused by the restricted field of view of a single camera, and providing complete trajectories of targets moving across different regions. However, several challenges remain in practical applications. First, differences in camera viewpoints, illumination conditions, and image resolutions often cause significant appearance variations of the same target across different cameras, making cross-camera identity association more difficult. Second, target occlusion, temporary disappearance from the camera view, and interactions among multiple targets can easily lead to tracking interruptions or identity switches, thereby degrading the overall tracking performance.
To address these challenges and improve the stability and accuracy of multi-camera target tracking, this study focuses on three key aspects: (1) spatial geometric mapping on a shared ground plane, (2) cross-camera trajectory association based on movement direction and spatial distance, and (3) cross-temporal identity consistency maintenance using historical identity information. To overcome the challenges of cross-camera identity association, a two-stage cross-camera matching framework is proposed. The framework first establishes initial associations by exploiting geometric information on the shared ground plane together with target motion characteristics, and then resolves identity conflicts and unifies global identities through historical identity information and a group consistency strategy.
In the experimental evaluation, a multi-camera pedestrian tracking dataset generated in a virtual environment is adopted to assess the proposed method, and multiple evaluation metrics are used to analyze its tracking performance. The case studies demonstrate that the proposed approach effectively maintains stable cross-camera associations by utilizing motion information and, in certain scenarios, even outperforms several existing methods. The experimental results indicate that integrating shared-ground-plane geometric information with motion direction features can achieve stable and competitive tracking performance while maintaining low computational cost. Furthermore, the proposed method establishes reliable target associations without relying on high-dimensional appearance features, thereby improving global identity matching accuracy and the overall tracking quality.

中文摘要	 I
Abstract	 II
List of Tables	 VI
List of Figures	 VII
Chapter 1  Introduction	 1
1  Background	 1
2  Motivation  2
3  Research Approach  3
4  Contribution  5
5  Organization  6
Chapter 2  Related Work  7
1  Object Detection Algorithm  8
2  Single-Camera Multi-Object Tracking, SC-MOT	 10
3  Multi-Camera Multi-Object Tracking, MC-MOT  14
3.1  Cross-Camera Association Methods Based on Clustering Strategies  15
3.2  Non-Clustering-Based Cross-Camera Association Methods  21
Chapter 3  Problem Statement  26
Chapter 4  Methodology  29
1  Detection  31
2  Single-Camera Multi-Object Tracking  32
3  Multi-Camera Multi-Object Tracking  34
4  Homography Matrix  35
5  Cross-Camera Multi-Object Tracking  39
5.1  The First Merge ─ The Angle and Distance Between Vectors	 40
5.2  Exclude Duplicate Global IDs  42
5.3  Cross-Time Connection  44
5.4  The Second Merge ─ History Trajectory Vector Similarity  48
Chapter 5  Experimental Evaluation  51
1  Experimental Data, Setting and Evaluation Metrics  51
1.1  Experimental Data and Setting  52
1.2  Evaluation Metrics  53
2  Internal Experiment  56
2.1  Vector Angle and Distance  57
2.2  Internal Parameter and Performance Analysis  62
2.3  Comparison of Different Single-Camera Trackers  68
3  External Experiment  71
4  Case Study  85
Chapter 6  Conclusions and Future Work  92
References  96
                                    

[1] Aharon, N., Orfaig, R., & Bobrovsky, B. Z. (2022). BoT-SORT: Robust associations multi-pedestrian tracking. arXiv preprint arXiv:2206.14651.
[2] Bewley, A., Ge, Z., Ott, L., Ramos, F., & Upcroft, B. (2016, September). Simple online and realtime tracking. In 2016 IEEE international conference on image processing (ICIP) (pp. 3464-3468). IEEE.
[3] Cherdchusakulchai, R., Phimsiri, S., Trairattanapa, V., Tungjitnob, S., Kudisthalert, W., Kiawjak, P., ... & Suttichaya, V. (2024). Online Multi-camera People Tracking with Spatial-temporal Mechanism and Anchor-feature Hierarchical Clustering. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (pp. 7198-7207).
[4] Girshick, R. (2015). Fask r-cnn. In Proceedings of the IEEE international conference on computer vision (pp. 1440-1448).
[5] He, K., Gkioxari, G., Dollár, P., & Girshick, R. (2017). Mask r-cnn. In Proceedings of the IEEE international conference on computer vision (pp. 2961-2969).
[6] He, Y., Wei, X., Hong, X., Shi, W., & Gong, Y. (2020). Multi-target multi-camera tracking by tracklet-to-target assignment. IEEE Transactions on Image Processing, 29, 5191-5205.
[7] Hou, Y., Wang, Z., Wang, S., & Zheng, L. (2021). Adaptive affinity for associations in multi-target multi-camera tracking. IEEE Transactions on Image Processing, 31, 612-622.
[8] Huang, H. W., Yang, C. Y., Jiang, Z., Kim, P. K., Lee, K., Kim, K., ... & Hwang, J. N. (2023). Enhancing multi-camera people tracking with anchor-guided clustering and spatio-temporal consistency id re-assignment. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (pp. 5239-5249).
[9] Jang, J., Seon, M., & Choi, J. (2022). Lightweight indoor multi-object tracking in overlapping FOV multi-camera environments. Sensors, 22(14), 5267.
[10] Kim, J., Shin, W., Park, H., & Choi, D. (2024). Cluster self-refinement for enhanced online multi-camera people tracking. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (pp. 7190-7197).
[11] Kohl, P., Specker, A., Schumann, A., & Beyerer, J. (2020). The mta dataset for multi-target multi-camera pedestrian tracking by weighted distance aggregation. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops (pp. 1042-1043).
[12] Luiten, J., Osep, A., Dendorfer, P., Torr, P., Geiger, A., Leal-Taixé, L., & Leibe, B. (2021). Hota: A higher order metric for evaluating multi-object tracking. International journal of computer vision, 129(2), 548-578.
[13] Phan, T. H., Dinh, D. D., Huynh, T. P., Le, Q. T., Dang, H. H., Tran, V. H., ... & Huang, C. C. (2025). Vgcrtrack: Multi-camera 3d tracking with view-aware geometric center refinement. In Proceedings of the IEEE/CVF International Conference on Computer Vision (pp. 5375-5381).
[14] Quach, K. G., Nguyen, P., Le, H., Truong, T. D., Duong, C. N., Tran, M. T., & Luu, K. (2021). Dyglip: A dynamic graph model with link prediction for accurate multi-camera multiple object tracking. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (pp. 13784-13793).
[15] Qian, Y., Yu, L., Liu, W., & Hauptmann, A. (2020, June). Electricity: An efficient multi-camera vehicle tracking system for intelligent city. 2020 IEEE. In CVF Conference on Computer Vision and Pattern Recognition Workshops (CVPRW) (pp. 2511-2519).
[16] Shuo Wang, David C Anastasiu, Zheng Tang, MingChing Chang, Yue Yao, Liang Zheng, Mohammed Shaiqur Rahman, Meenakshi S Arya, Anuj Sharma, Pranamesh Chakraborty, et al. The 8th ai city challenge. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 7261–7272, 2024. 2, 5, 6
[17] Specker, A. (2024). OCMCTrack: Online multi-target multi-camera tracking with corrective matching cascade. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (pp. 7236-7244).
[18] Tran, L. Q., & Vi, H. D. (2024). Efficient online multi-camera tracking with memory-efficient accumulated appearance features and trajectory validation. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (pp. 7217-7226).
[19] Tran, T. H. P., Tran, D. N. N., Huynh, N. D. M., Tran, C. D., Pham, L. H., Ho, Q. P. N., ... & Jeon, J. W. (2025). Depthtrack: Cluster meets bev for multi-camera multi-target 3d tracking. In Proceedings of the IEEE/CVF International Conference on Computer Vision (pp. 5289-5298).
[20] Varghese, R., & Sambath, M. (2024, April). Yolov8: A novel object detection algorithm with enhanced performance and robustness. In 2024 International conference on advances in data engineering and intelligent computing systems (ADICS) (pp. 1-6). IEEE.
[21] Wang, Y., Meinhardt, T., Cetintas, O., Yang, C. Y., Pusegaonkar, S., Missaoui, B., ... & Leal-Taixe, L. (2025). MCBLT: Multi-camera multi-object 3D tracking in long videos. In Proceedings of the IEEE/CVF International Conference on Computer Vision (pp. 5245-5254).
[22] Wang, Z., Zheng, L., Liu, Y., Li, Y., & Wang, S. (2020, August). Towards real-time multi-object tracking. In European conference on computer vision (pp. 107-122). Cham: Springer International Publishing.
[23] Wojke, N., Bewley, A., & Paulus, D. (2017, September). Simple online and real-time tracking with a deep association metric. In 2017 IEEE international conference on image processing (ICIP) (pp. 3645-3649). IEEE.
[24] Wu, M., Qian, Y., Wang, C., & Yang, M. (2021). A multi-camera vehicle tracking system based on city-scale vehicle re-id and spatial-temporal information. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (pp. 4077-4086).
[25] Xie, Z., Ni, Z., Yang, W., Zhang, Y., Chen, Y., Zhang, Y., & Ma, X. (2024). A Robust Online Multi-Camera People Tracking System With Geometric Consistency and State-aware Re-ID Correction. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (pp. 7007-7016).
[26] Yang, W., Xie, Z., Wang, Y., Zhang, Y., Ma, X., & Hao, B. (2023). Integrating appearance and spatial-temporal information for multi-camera people tracking. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (pp. 5260-5269).
[27] Zhang, R., Wu, L., Yang, Y., Wu, W., Chen, Y., & Xu, M. (2020). Multi-camera multi-player tracking with deep player identification in sports video. Pattern Recognition, 102, 107260.
[28] Zhang, Y., Sun, P., Jiang, Y., Yu, D., Weng, F., Yuan, Z., ... & Wang, X. (2022, October). Bytetrack: Multi-object tracking by associating every detection box. In European conference on computer vision (pp. 1-21). Cham: Springer Nature Switzerland.

簡易檢索 / 詳目顯示

相關論文