簡易檢索 / 詳目顯示

回結果列表
研究生： 劉玠旼
Liu, Chieh-Min
論文名稱： 高仿真度微觀車流分析：結合深度學習與時間序列追蹤之多平台方法
High-Fidelity Microscopic Traffic Flow Analysis: A Multi-Platform Approach Using Deep Learning and Temporal Tracking
指導教授： 莊智清
Juang, Jyh-Ching
學位類別： 博士
Doctor
系所名稱： 電機資訊學院 - 電機工程學系
Department of Electrical Engineering
論文出版年： 2026
畢業學年度： 114
語文別： 英文
論文頁數： 128
中文關鍵詞： 智慧交通系統 、多目標追蹤 、深度學習 、無人機 、YOLO 、DeepSORT 、微觀車流
外文關鍵詞： Intelligent Transportation Systems (ITS), Multi-Object Tracking, Deep Learning, Unmanned Aerial Vehicles (UAVs), YOLO, DeepSORT, Microscopic Traffic Flow
相關次數： 點閱：18  下載：0 
分享至:
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

    • 精準感知微觀車流為現代智慧交通系統之重要基石。傳統的感測器（如電子收費系統 ETC）高度依賴實體基礎設施，且常面臨監控盲區與高昂建置成本等限制。為突破上述瓶頸，本論文提出一套綜合性之多平台框架，透過整合固定門架攝影機與無人機，以實現高仿真度之車道級車流估計。
    首先，針對地面連續影像中常見的軌跡不一致問題，本研究開發了一套整合 YOLOv4、DeepSORT 以及創新「虛擬線與熱區」邏輯機制的視覺追蹤系統。此方法有效克服了軌跡破碎與車輛重複計數之缺點，從而精準計算出各車道之車流相關指標與車輛個別車速。
    其次，將視角轉移至無基礎設施限制的空中監控，本研究系統性地評估了 YOLOv8 架構家族，為無人機影像中嚴重的像素稀缺與極端尺度變化，建立量化基準與縮放指南。基於 VisDrone 資料集的量化評估結果打破了深度學習中「模型越大越好」的傳統迷思，證實單純提升輸入解析度（YOLOv8l @ 1280）即可帶來相較 YOLOv5 基準 323% 的巨大性能提升，其成效遠勝於複雜的網路架構修改。
    最後，為整合高解析度之空間偵測與時間序列追蹤，本論文提出了一套以無人機為核心之多平台監控架構藍圖。除透過影像對齊以補償無人機之自我移動、運用透視變換將動態虛擬線映射至路面，並採用「邊緣到雲端」(Edge-to-Cloud, E2C)架構以解決邊緣設備的運算瓶頸外，本研究進一步將此移動式系統與國道既有之感測基礎設施（固定式攝影機、車輛偵測器與電子收費門架）整合，形成多平台之智慧交通感測網路。在此架構下，該框架於理論層面上實現了基於移動空拍視角的動態多目標追蹤。
    總結而言，本論文提供了充分的理論見解與具實證基礎的系統部署指南，為全面性的微觀交通監控提供了一套具備高度擴充性且具成本效益的視覺感測解決方案，並為未來的自駕車導航與高速公路交通管理奠定堅實的基礎。

    Accurate perception of microscopic traffic flow is a fundamental pillar of modern Intelligent Transportation Systems (ITS). Traditional infrastructure-dependent sensors, such as Electronic Toll Collection (ETC) systems, suffer from monitoring blind spots and high installation costs. To address these limitations, this dissertation proposes a comprehensive, multi-platform framework that integrates fixed-gantry cameras and Unmanned Aerial Vehicles (UAVs) to achieve high-fidelity, lane-level traffic flow estimation.
    First, to resolve trajectory inconsistencies in ground-level continuous video streams, a vision-based tracking system integrating YOLOv4, DeepSORT, and a novel logical mechanism of "virtual lines" and "hot zones" is developed. This methodology effectively prevents trajectory fragmentation and the double-counting of identical vehicles, enabling the highly accurate estimation of lane-specific flow-related metrics and individual velocities.
    Second, shifting to the infrastructure-less aerial perspective, this research systematically evaluates the YOLOv8 architecture family to establish quantitative baselines and scaling guidelines for small-object detection under severe pixel scarcity and extreme scale variations inherent in drone imagery. Quantitative evaluations on the highly cluttered VisDrone dataset challenge the prevailing "bigger is always better" paradigm. The results demonstrate that straightforward input resolution scaling (YOLOv8l at 1280) yields a remarkable 323% accuracy improvement over the YOLOv5 baseline, vastly outperforming complex architectural modifications, thereby providing a highly cost-effective spatial sensing foundation before introducing bespoke modules.
    Finally, to integrate high-resolution spatial detection with temporal tracking, this dissertation proposes a UAV-centered, multi-platform monitoring blueprint. Beyond incorporating image registration to compensate for UAV ego-motion, perspective transformation to map dynamic virtual lines onto the road plane, and an Edge-to-Cloud (E2C) computing architecture to resolve severe computational bottlenecks on edge devices, the framework further coordinates this mobile system with existing freeway sensing infrastructure, such as fixed cameras, vehicle detectors, and eTag gantries, to form a heterogeneous, multi-platform ITS sensor network. Within this architecture, the framework theoretically realizes dynamic multi-object tracking from moving aerial viewpoints.
    Ultimately, this dissertation provides both valuable theoretical insights and evidence-based deployment guidelines. It delivers a highly scalable and cost-effective vision-based sensing solution for comprehensive microscopic traffic monitoring, paving the way for future advancements in autonomous navigation and freeway traffic management.

    List of Tables xiv List of Figures xv Chapter 1 Introduction 1 1.1 Background and Motivation 1 1.2 Problem Statement and Challenges 2 1.2.1 Challenge 1: Trajectory Inconsistencies in Fixed-Camera Tracking 3 1.2.2 Challenge 2: Severe Pixel Scarcity and Scale Variations in Aerial Imagery 3 1.3 Research Objectives 5 1.4 Main Contributions 6 1.4.1 Innovations in Microscopic Traffic Tracking (Fixed-Gantry Platform) 6 1.4.2 Scaling Strategies for Aerial Sensing (UAV Platform) 7 1.4.3 Integrating UAV Sensing, Fixed-Camera and Existing Freeway Sensing Infrastructure for Multi-Platform Monitoring 8 1.5 Dissertation Organization 9 Chapter 2 Theoretical Foundations of Microscopic Traffic Models and Computer Vision 11 2.1 Prior Work on Object Detection 11 2.1.1 General Object Detection 11 2.1.2 Small-Object Detection 12 2.1.3 Drone-Based Object Detection 15 2.2 The Evolution of YOLO Architecture Series 17 2.2.1 The Paradigm Shift to One-Stage Detection (YOLOv1 ~ YOLOv3) 17 2.2.2 YOLOv4: Optimizing Speed and Accuracy for Real-Time Deployment 19 2.2.3 YOLOv5 to YOLOv8: Compound Scaling and Anchor-Free Paradigms 20 2.3 Multi-Object Tracking and DeepSORT 23 2.4 Microscopic and Macroscopic Traffic Flow Models 26 2.5 Summary and Research Gaps 28 Chapter 3 Estimating Vehicle Velocity and Lane-Specific Density via YOLOv4 and DeepSORT Integration 30 3.1 Experimental Methods 30 3.1.1 Overall Experimental Process 30 3.1.2 Pretrain and Fine-Tuning for Vehicle Identification 30 3.1.3 Multi-Object Tracking Using Virtual Lines and Hot Zones 34 3.1.4 Estimation of Velocities of Each Vehicle 35 3.2 Experimental Results 37 3.2.1 Digital Image Processing 37 3.2.2 Object Detection Results 37 3.2.3 Ablation Study on the Hot Zone Mechanism 40 3.2.4 Vehicle Counting in Both Directions 42 3.2.5 Lane-Specific Vehicle Counting 43 3.2.6 Velocity Estimation 45 3.2.7 Velocity Level Visualization 46 3.3 Discussion 47 3.4 Limitations 48 Chapter 4 Scaling Strategies and Quantitative Baselines for Small-Object Detection in Densely Cluttered Urban Aerial Scenes 50 4.1 Small-Object Detection in Drone Imagery 50 4.1.1 Challenges in Aerial Detection and Evaluation Strategy 50 4.1.2 Experimental Setup 52 4.2 Experiments 54 4.2.1 Model Capacity Effect 54 4.2.2 Input Resolution Effect 55 4.2.3 Failure Mechanism of Oversized Models 56 4.2.4 Architectural Modification Effect 58 4.2.5 Comprehensive Comparison 62 4.2.6 Speed-Accuracy Trade-Off Analysis 63 4.3 Discussion and Analysis 64 4.3.1 Performance Difference Between YOLOv8l and YOLO-HV 64 4.3.2 Per-Class Performance Analysis 67 4.3.3 Practical Implications 69 4.3.4 Conclusion on Optimal Configurations 72 4.3.5 Contextualizing Performance in Drone-Based Benchmarks 73 4.4 Summary and Limitations 73 Chapter 5 Discussions and Conceptual Blueprint for a UAV-Based, Multi-Platform Traffic Monitoring Framework 75 5.1 Introduction 75 5.2 Proposed UAV-Based System Architecture 76 5.2.1 Aerial Spatial Sensing Module 77 5.2.2 Temporal Tracking Module 78 5.2.3 Microscopic Traffic Analysis Module 78 5.3 Adapting Tracking Logic for UAV Ego-Motion 79 5.3.1 Overcoming Camera Motion in DeepSORT 79 5.3.2 Dynamic Virtual Lines via Perspective Transformation 80 5.3.3 Algorithmic Implementation of the Multi-Platform Framework 84 5.4 Deployment and Computational Feasibility 87 5.4.1 Computational Bottlenecks in UAV Edge Deployment 87 5.4.2 Edge-to-Cloud Offloading Strategy 88 5.5 System Feasibility Analysis and Testing Blueprint 89 5.5.1 Bandwidth Budget for E2C Offloading 89 5.5.2 Latency Budget and Hardware Extrapolation 89 5.5.3 Preliminary End-to-End Testing Plan 91 5.6 Integration with Existing Freeway Sensing Infrastructure 91 5.6.1 Complementary Sensor Modalities and Data Access 93 5.6.2 Cooperative Workflow 94 5.6.3 Reconciling Time-Mean and Space-Mean Speed 96 5.6.4 Common Spatial Reference and Data-Governance Preconditions 96 5.7 Summary 97 Chapter 6 Conclusions and Future Work 99 6.1 Summary of Main Findings 99 6.2 Key Academic and Practical Contributions 100 6.2.1 Academic Contributions 100 6.2.2 Practical Contributions 101 6.3 Discussions and Future Work 101 6.3.1 Limitations and Future Directions in Aerial Spatial Sensing 102 6.3.2 Limitations and Future Directions in Multi-Platform Traffic Monitoring 104 References 106

    [1] B. Tian, Q. Yao, Y. Gu, K. Wang, and Y. Li, “Video processing techniques for traffic flow monitoring: A survey,” in Proc. 14th Int. IEEE Conf. Intell. Transp. Syst. (ITSC), IEEE, 2011, pp. 1103–1108.
    [2] Y. Lv, Y. Duan, W. Kang, Z. Li, and F.-Y. Wang, “Traffic flow prediction with big data: A deep learning approach,” IEEE Trans. Intell. Transp. Syst., vol. 16, no. 2, pp. 865–873, 2014.
    [3] Freeway Bureau, Ministry of Transportation and Communications, “Freeway Bureau, MOTC Official Website,” Freeway Bureau,MOTC. Accessed: May 18, 2026. [Online]. Available: https://www.freeway.gov.tw/english/Default.aspx
    [4] C.-M. Liu and J.-C. Juang, “Estimation of lane-level traffic flow using a deep learning technique,” Appl. Sci., vol. 11, no. 12, p. 5619, 2021.
    [5] S. Srivastava, S. Narayan, and S. Mittal, “A survey of deep learning techniques for vehicle detection from UAV images,” J. Syst. Archit., vol. 117, p. 102152, 2021.
    [6] P. Zhu, L. Wen, X. Bian, H. Ling, and Q. Hu, “Vision meets drones: A challenge,” 2018, arXiv:1804.07437.
    [7] A. Milan, L. Leal-Taixé, I. Reid, S. Roth, and K. Schindler, “MOT16: A benchmark for multi-object tracking,” 2016, arXiv:1603.00831.
    [8] D. Du et al., “VisDrone-DET2019: The vision meets drone object detection in image challenge results,” in Proc. IEEE/CVF Int. Conf. Comput. Vis. Workshop, 2019, pp. 0–0.
    [9] D. Du et al., “The unmanned aerial vehicle benchmark: Object detection and tracking,” in Proc. Eur. Conf. Comput. Vis. (ECCV), 2018, pp. 370–386. Accessed: May 20, 2026. [Online]. Available: https://openaccess.thecvf.com/content_ECCV_2018/html/Dawei_Du_The_Unmanned_Aerial_ECCV_2018_paper.html
    [10] G.-S. Xia et al., “DOTA: A large-scale dataset for object detection in aerial images,” in Proc. IEEE/CVF Conf. Comput. Vis. Pattern Recognit., Salt Lake City, UT, USA, Jun. 2018, pp. 3974–3983. Accessed: May 19, 2026. [Online]. Available: https://ieeexplore.ieee.org/document/8578516/
    [11] A. Bochkovskiy, C.-Y. Wang, and H.-Y. M. Liao, “YOLOv4: Optimal speed and accuracy of object detection,” 2020, arXiv:2004.10934.
    [12] N. Wojke, A. Bewley, and D. Paulus, “Simple online and realtime tracking with a deep association metric,” in Proc. IEEE Int. Conf. Imag. Process. (ICIP), IEEE, 2017, pp. 3645–3649.
    [13] G. Jocher, A. Chaurasia, and J. Qiu, YOLOv8. (2023). Ultralytics. [Online]. Available: https://github.com/ultralytics/ultralytics
    [14] G. Jocher, YOLOv5. (2020). Ultralytics LLC. [Online]. Available: https://github.com/ultralytics/yolov5
    [15] B. Seibold, “A mathematical introduction to traffic flow theory,” Institute for Pure and Applied Mathematics (IPAM), Univ. of California, Los Angeles, CA, USA, 2015.
    [16] R. Girshick, J. Donahue, T. Darrell, and J. Malik, “Rich feature hierarchies for accurate object detection and semantic segmentation,” in Proc. IEEE Conf. Comput. Vis. Pattern Recognit., Columbus, OH, USA, Jun. 2014, pp. 580–587. doi: 10.1109/CVPR.2014.81.
    [17] J. Redmon, S. Divvala, R. Girshick, and A. Farhadi, “You only look once: Unified, real-time object detection,” in Proc. IEEE Conf. Comput. Vis. Pattern Recognit., 2016, pp. 779–788.
    [18] W. Liu et al., “SSD: Single shot MultiBox detector,” in Proc. Eur. Conf. Comput. Vis. (ECCV), 2016, pp. 21–37. doi: 10.1007/978-3-319-46448-0_2.
    [19] T.-Y. Lin, P. Goyal, R. Girshick, K. He, and P. Dollar, “Focal loss for dense object detection,” IEEE Trans. Pattern Anal. Mach. Intell., vol. 42, no. 2, pp. 318–327, Feb. 2020, doi: 10.1109/TPAMI.2018.2858826.
    [20] M. Tan, R. Pang, and Q. V. Le, “EfficientDet: Scalable and efficient object detection,” in Proc. IEEE/CVF Conf. Comput. Vis. Pattern Recognit. (CVPR), Seattle, WA, USA: IEEE, Jun. 2020, pp. 10778–10787. doi: 10.1109/CVPR42600.2020.01079.
    [21] S. Zhang, C. Chi, Y. Yao, Z. Lei, and S. Z. Li, “Bridging the gap between anchor-based and anchor-free detection via adaptive training sample selection,” in Proc. IEEE/CVF Conf. Comput. Vis. Pattern Recognit. (CVPR), Seattle, WA, USA, Jun. 2020, pp. 9756–9765. doi: 10.1109/CVPR42600.2020.00978.
    [22] T.-Y. Lin et al., “Microsoft COCO: Common objects in context,” in Proc. Eur. Conf. Comput. Vis. (ECCV), 2014, pp. 740–755. doi: 10.1007/978-3-319-10602-1_48.
    [23] T.-Y. Lin, P. Dollár, R. Girshick, K. He, B. Hariharan, and S. Belongie, “Feature pyramid networks for object detection,” in Proc. IEEE Conf. Comput. Vis. Pattern Recognit., 2017, pp. 2117–2125.
    [24] B. Singh and L. S. Davis, “An analysis of scale invariance in object detection snip,” in Proc. IEEE Conf. Comput. Vis. Pattern Recognit., 2018, pp. 3578–3587.
    [25] S. Liu, L. Qi, H. Qin, J. Shi, and J. Jia, “Path aggregation network for instance segmentation,” in Proc. IEEE/CVF Conf. Comput. Vis. Pattern Recognit., Salt Lake City, UT, USA, Jun. 2018, pp. 8759–8768. doi: 10.1109/CVPR.2018.00913.
    [26] G. Ghiasi, T.-Y. Lin, and Q. V. Le, “NAS-FPN: Learning scalable feature pyramid architecture for object detection,” in Proc. IEEE/CVF Conf. Comput. Vis. Pattern Recognit. (CVPR), Long Beach, CA, USA, Jun. 2019, pp. 7029–7038. doi: 10.1109/CVPR.2019.00720.
    [27] S. Woo, J. Park, J.-Y. Lee, and I. S. Kweon, “CBAM: Convolutional block attention module,” in Proc. Eur. Conf. Comput. Vis. (ECCV), 2018, pp. 3–19.
    [28] Q. Wang, B. Wu, P. Zhu, P. Li, W. Zuo, and Q. Hu, “ECA-Net: Efficient channel attention for deep convolutional neural networks,” in Proc. IEEE/CVF Conf. Comput. Vis. Pattern Recognit. (CVPR), Seattle, WA, USA: IEEE, Jun. 2020, pp. 11531–11539. doi: 10.1109/CVPR42600.2020.01155.
    [29] X. Wang, R. Girshick, A. Gupta, and K. He, “Non-local neural networks,” in Proc. IEEE/CVF Conf. Comput. Vis. Pattern Recognit., Salt Lake City, UT, USA, Jun. 2018, pp. 7794–7803. doi: 10.1109/CVPR.2018.00813.
    [30] Z. Liu et al., “Swin Transformer: Hierarchical vision transformer using shifted windows,” in Proc. IEEE/CVF Int. Conf. Comput. Vis. (ICCV), Montreal, QC, Canada, Oct. 2021, pp. 9992–10002. doi: 10.1109/ICCV48922.2021.00986.
    [31] J. Dai et al., “Deformable convolutional networks,” in Proc. IEEE Int. Conf. Comput. Vis. (ICCV), Venice, Italy, Oct. 2017, pp. 764–773. doi: 10.1109/ICCV.2017.89.
    [32] X. Zhu, W. Su, L. Lu, B. Li, X. Wang, and J. Dai, “Deformable DETR: Deformable transformers for end-to-end object detection,” in Int. Conf. Learn. Represent., Oct. 2020. Accessed: May 20, 2026. [Online]. Available: https://openreview.net/forum?id=gZ9hCDWe6ke
    [33] J. Wang, K. Chen, R. Xu, Z. Liu, C. C. Loy, and D. Lin, “CARAFE: Content-Aware ReAssembly of FEatures,” in Proc. IEEE/CVF Int. Conf. Comput. Vis., 2019, pp. 3007–3016.
    [34] W. Liu, H. Lu, H. Fu, and Z. Cao, “Learning to upsample by learning to sample,” in Proc. IEEE/CVF Int. Conf. Comput. Vis. (ICCV), Paris, France, Oct. 2023, pp. 6004–6014. doi: 10.1109/ICCV51070.2023.00554.
    [35] G. Ghiasi et al., “Simple copy-paste is a strong data augmentation method for instance segmentation,” in Proc. IEEE/CVF Conf. Comput. Vis. Pattern Recognit. (CVPR), Nashville, TN, USA, Jun. 2021, pp. 2917–2927. doi: 10.1109/CVPR46437.2021.00294.
    [36] Z. Cai and N. Vasconcelos, “Cascade R-CNN: Delving into high quality object detection,” in Proc. IEEE/CVF Conf. Comput. Vis. Pattern Recognit., Salt Lake City, UT, Jun. 2018, pp. 6154–6162. doi: 10.1109/CVPR.2018.00644.
    [37] X. Zhu, S. Lyu, X. Wang, and Q. Zhao, “TPH-YOLOv5: Improved YOLOv5 based on transformer prediction head for object detection on drone-captured scenarios,” in Proc. IEEE/CVF Int. Conf. Comput. Vis. Workshops (ICCVW), Montreal, BC, Canada, Oct. 2021, pp. 2778–2788. doi: 10.1109/ICCVW54120.2021.00312.
    [38] S. Xu, M. Zhang, J. Chen, and Y. Zhong, “YOLO-HyperVision: A vision transformer backbone-based enhancement of YOLOv5 for detection of dynamic traffic information,” Egypt. Inform. J., vol. 27, p. 100523, 2024.
    [39] B. Singh, M. Najibi, and L. S. Davis, “Sniper: Efficient multi-scale training,” Adv. Neural Inf. Process. Syst., vol. 31, 2018.
    [40] J. Liao and H. Tian, “Cluster-NMS: Improving crowded object detection through clustering pattern,” Signal, Image Video Process., vol. 19, no. 9, pp. 1–8, 2025.
    [41] J. Redmon and A. Farhadi, “YOLOv3: An incremental improvement,” 2018, arXiv:1804.02767. [Online]. Available: http://arxiv.org/abs/1804.02767
    [42] C.-Y. Wang, H.-Y. M. Liao, Y.-H. Wu, P.-Y. Chen, J.-W. Hsieh, and I.-H. Yeh, “CSPNet: A new backbone that can enhance learning capability of CNN,” in Proc. IEEE/CVF Conf. Comput. Vis. Pattern Recognit., 2020, pp. 390–391.
    [43] C. Li et al., “YOLOv6: A single-stage object detection framework for industrial applications,” 2022, arXiv:2209.02976. Accessed: May 19, 2026. [Online]. Available: https://arxiv.org/abs/2209.02976v1
    [44] C.-Y. Wang, A. Bochkovskiy, and H.-Y. M. Liao, “YOLOv7: Trainable bag-of-freebies sets new state-of-the-art for real-time object detectors,” in Proc. IEEE/CVF Conf. Comput. Vis. Pattern Recognit. (CVPR), Jun. 2023, pp. 7464–7475. doi: 10.1109/CVPR52729.2023.00721.
    [45] G. Ciaparrone, F. L. Sánchez, S. Tabik, L. Troiano, R. Tagliaferri, and F. Herrera, “Deep learning in video multi-object tracking: A survey,” Neurocomputing, vol. 381, pp. 61–88, 2020.
    [46] A. Bewley, Z. Ge, L. Ott, F. Ramos, and B. Upcroft, “Simple online and realtime tracking,” in Proc. IEEE Int. Conf. Imag. Process. (ICIP), 2016, pp. 3464–3468.
    [47] Z. Wang, L. Zheng, Y. Liu, Y. Li, and S. Wang, “Towards real-time multi-object tracking,” in Proc. Eur. Conf. Comput. Vis. (ECCV), 2020, pp. 107–122.
    [48] A. Fedorov, K. Nikolskaia, S. Ivanov, V. Shepelev, and A. Minbaleev, “Traffic flow estimation with data from a video surveillance camera,” J. Big Data, vol. 6, no. 1, p. 73, 2019.
    [49] A. M. Santos, C. J. Bastos-Filho, A. M. Maciel, and E. Lima, “Counting vehicle with high-precision in brazilian roads using yolov3 and deep sort,” in Proc. 33rd SIBGRAPI Conf. Graph. Patterns Imag. (SIBGRAPI), 2020, pp. 69–76.
    [50] N. S. Punn, S. K. Sonbhadra, S. Agarwal, and G. Rai, “Monitoring COVID-19 social distancing with person detection and tracking via fine-tuned YOLO v3 and Deepsort techniques,” 2020, arXiv:2005.01385.
    [51] Z. Qiu et al., “Vision-based moving obstacle detection and tracking in paddy field using improved yolov3 and deep SORT,” Sensors, vol. 20, no. 15, p. 4082, 2020.
    [52] Z. Tian, C. Shen, H. Chen, and T. He, “Fcos: Fully convolutional one-stage object detection,” in Proc. IEEE/CVF Int. Conf. Comput. Vis., 2019, pp. 9627–9636.
    [53] X. Li et al., “Generalized focal loss: Learning qualified and distributed bounding boxes for dense object detection,” in Adv. Neural Inf. Process. Syst., 2020, pp. 21002–21012. Accessed: May 20, 2026. [Online]. Available: https://proceedings.neurips.cc/paper/2020/hash/f0bda020d2470f2e74990a07a607ebd9-Abstract.html
    [54] C.-M. Liu and J.-C. Juang, “Systematic evaluation of YOLOv8 variants for UAV-based object detection,” Appl. Sci., vol. 16, no. 7, p. 3559, Apr. 2026, doi: 10.3390/app16073559.
    [55] Y. Lin, Y. Lin, H. Wu, and M. Wu, “The enhance-fuse-align principle: A new architectural blueprint for robust object detection, with application to X-ray security,” Sensors, vol. 25, no. 21, p. 6603, 2025.
    [56] A. C. Shastry and R. A. Schowengerdt, “Airborne video registration and traffic-flow parameter estimation,” IEEE Trans. Intell. Transp. Syst., vol. 6, no. 4, pp. 391–405, 2005.
    [57] “TDX | TDX 運輸資料流通服務.” Accessed: Jun. 22, 2026. [Online]. Available: https://tdx.transportdata.tw/
    [58] N.-E. E. Faouzi, H. Leung, and A. Kurian, “Data fusion in intelligent transportation systems: Progress and challenges – A survey,” Inf. Fusion, vol. 12, no. 1, pp. 4–10, Jan. 2011, doi: 10.1016/j.inffus.2010.06.001.
    [59] J. G. Wardrop, “Road paper. Some theoretical aspects of road traffic research,” Proc. Inst. Civil Eng., vol. 1, no. 3, pp. 325–362, May 1952, doi: 10.1680/ipeds.1952.11259.

    下載圖示
    校外：立即公開
    QR CODE