隨著城市交通的增長，交通事故和擁堵問題成為城市管理者亟需解決的難題。根據世界衛生組織報告，每年全球約有135萬人因交通事故喪生，其中行人，特別是輪椅使用者等行動不便的群體，佔很大比例。提升城市路口的安全性和交通流暢度至關重要。為此，本研究提出了一種基於基礎語言圖像預訓練的即時特殊行人檢測技術，並將其與智慧交通號誌控制系統融合，以優化城市路口的安全性和交通流暢度。

在前端處理方面，我們使用了 YOLOv7 模型進行行人檢測，這是一種高效的實時物體檢測模型。後處理則採用了 GLIP 技術，結合自然語言處理和圖像處理的優勢，進一步提高行人檢測的精度和可靠性，特別是對輪椅使用者等特殊行人的檢測。智慧交通號誌控制系統通過檢測和分析行人數據動態調整交通信號的秒數，確保行人能夠安全通過馬路。

本研究提供了一種創新的解決方案，不僅提高了城市路口的安全性和交通流暢度，還為未來智慧城市的發展提供了重要參考。隨著技術的進一步完善和推廣，這一系統有望在全球範圍內得到廣泛應用，實現更加安全、高效的城市交通管理。

With the growth of urban traffic, traffic accidents and congestion have become pressing issues for city administrators. According to the World Health Organization, approximately 1.35 million people worldwide die in traffic accidents each year, with a significant proportion involving pedestrians, especially those with mobility impairments such as wheelchair users. Enhancing the safety and efficiency of urban intersections is crucial. To address this challenge, this study proposes a real-time special pedestrian detection technology based on grounded language-image pre-training, integrated with intelligent traffic signal control systems to optimize urban intersection safety and traffic flow.
In terms of frontend processing, we employ the YOLOv7 model for pedestrian detection, a highly efficient real-time object detection model. For post-processing, the GLIP (Grounded Language-Image Pre-training) technique is utilized, leveraging the advantages of natural language processing and image processing to further improve the accuracy and reliability of pedestrian detection, particularly for special pedestrians like wheelchair users. The intelligent traffic signal control system dynamically adjusts signal timings based on pedestrian data analysis to ensure safe pedestrian passage.
This study provides an innovative solution that not only enhances urban intersection safety and traffic flow but also serves as an important reference for the future development of smart cities. With further refinement and promotion of the technology, this system is poised to be widely adopted globally, leading to safer and more efficient urban traffic management.

摘要....i
ABSTRACT.......ii
誌謝........iii
目錄......iv
表目錄......vi
圖目錄......vii
第一章序論...... 1
1.1研究背景與動機......1
1.2研究方法......3
1.3文獻回顧與探討......4
1.4論文組織......12
第二章背景知識......13
2.1語意模型介紹......13
2.2實測模型比較......18
2.3YOLO影像辨識......20
第三章系統設計流程......23
3.1特殊行人資料集......23
3.2軟體流程......25
3.3設備流程架構......26
3.4行人穿越影像測試資料來源......27
3.5實驗環境與工具......27
第四章實驗結果......29
4.1模型評估 - 混淆矩陣......29
4.2特殊行人資料集訓練成果......31
4.3 YOLO偵測......32
4.4 GLIP與YOLO WORD......34
4.5 YOLO偵測與GLIP偵測......36
4.6道路實際測試......38
4.7 網路實景圖檢測準確率......40
第五章結論與未來展望......45
結論......45
未來展望......45
REFERENCE......46
專有名詞對照......49

[1]W. H. Organization, Global status report on road safety 2023: summary. World Health Organization, 2023.
[2]交通部路政及道安司. "行人交通事故死亡統計." 交通部. https://roadsafety.tw/ (accessed 2024).
[3]L. H. Li et al., "Grounded language-image pre-training," in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2022, pp. 10965-10975.
[4]Wang, Chien-Yao, I-Hau Yeh, and Hong-Yuan Mark Liao. "Yolov9: Learning what you want to learn using programmable gradient information." arXiv preprint arXiv:2402.13616 (2024).
[5]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 Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, 2023, pp. 7464-7475.
[6]M. Sharma, A. Bansal, V. Kashyap, P. Goyal, and T. H. Sheikh, "Intelligent traffic light control system based on traffic environment using deep learning," in IOP Conference Series: Materials Science and Engineering, 2021, vol. 1022, no. 1: IOP Publishing, p. 012122.
[7]J. Redmon and A. Farhadi, "Yolov3: An incremental improvement," arXiv preprint arXiv:1804.02767, 2018.
[8]M. M. Gandhi, D. S. Solanki, R. S. Daptardar, and N. S. Baloorkar, "Smart control of traffic light using artificial intelligence," in 2020 5th IEEE international conference on recent advances and innovations in engineering (ICRAIE), 2020: IEEE, pp. 1-6
[9]Y. Pi, N. Duffield, A. H. Behzadan, and T. Lomax, "Visual recognition for urban traffic data retrieval and analysis in major events using convolutional neural networks," Computational urban science, vol. 2, no. 1, p. 2, 2022.
[10]M. Sha and A. Boukerche, "Performance evaluation of CNN-based pedestrian detectors for autonomous vehicles," Ad Hoc Networks, vol. 128, p. 102784, 2022
[11]T.-T. Dao, Q.-V. Pham, and W.-J. Hwang, "FastMDE: A fast CNN architecture for monocular depth estimation at high resolution," IEEE Access, vol. 10, pp. 16111-16122, 2022.
[12]S. Iftikhar, Z. Zhang, M. Asim, A. Muthanna, A. Koucheryavy, and A. A. Abd El-Latif, "Deep learning-based pedestrian detection in autonomous vehicles: Substantial issues and challenges," Electronics, vol. 11, no. 21, p. 3551, 2022.
[13]B. Ghari, A. Tourani, and A. Shahbahrami, "A robust pedestrian detection approach for autonomous vehicles," in 2022 8th Iranian Conference on Signal Processing and Intelligent Systems (ICSPIS), 2022: IEEE, pp. 1-5.
[14]T. Cheng, L. Song, Y. Ge, W. Liu, X. Wang, and Y. Shan, "Yolo-world: Real-time open-vocabulary object detection," in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2024, pp. 16901-16911.
[15]A. Radford et al., "Learning transferable visual models from natural language supervision," in International conference on machine learning, 2021: PMLR, pp. 8748-8763.
[16]T. Lin, M. Maire, S. Belongie, L. Bourdev, R. Girshick, J. Hays, P. Perona, D. Ramanan, P. Dollar, and C. L. Zitnick, "Microsoft COCO: Common Objects in Context," in Proceedings of the European Conference on Computer Vision (ECCV), Zurich, Switzerland, Sep. 2014, pp. 740-755. doi: 10.1007/978-3-319-10602-1_48..
[17]P. Shao, Y. Zhu, Z. Zhao, B. Li, G. Zhao, L. Yu, T. Zhao, X. Liu, and J. Jia, "Objects365: A Large-scale, High-quality Dataset for Object Detection," in Proceedings of the IEEE International Conference on Computer Vision (ICCV), Seoul, Korea, Oct. 2019, pp. 8430-8439. doi: 10.1109/ICCV.2019.00870.
[18]A. Gupta, P. Dollar, and R. Girshick, "LVIS: A Dataset for Large Vocabulary Instance Segmentation," in Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR), Long Beach, CA, USA, Jun. 2019, pp. 5356-5364. doi: 10.1109/CVPR.2019.00550.
[19]S. Liu et al., "Grounding dino: Marrying dino with grounded pre-training for open-set object detection," arXiv preprint arXiv:2303.05499, 2023.
[20]L. Yao et al., "Detclip: Dictionary-enriched visual-concept paralleled pre-training for open-world detection," Advances in Neural Information Processing Systems, vol. 35, pp. 9125-9138, 2022.
[21]Hubery-Lee. "How to use python to crawl web images." https://github.com/Hubery-Lee/Notes (accessed 2024).
[22]Tzutalin. "label-studio." https://github.com/HumanSignal/labelImg (accessed 2024).
[23]M. M. Alani, "TCP/IP Model," in Guide to OSI and TCP/IP Models, SpringerBriefs in Computer Science. Cham: Springer, 2014, pp. 17-26. doi: 10.1007/978-3-319-05152-9_3).