臺灣博碩士論文加值系統

無人機提供了一種全面的視角來觀察環境，極大程度地避免了物件變形和遮擋問題，使其在監測交通及安全狀況方面具有顯著優勢。在本研究中，我們提出了一種基於無人機視角的動態人車衝突檢測模型，用於快速、準確地檢測道路中的潛在行人風險。為了提升檢測性能，我們對模型進行了改進，特別是增強了其對小物件的檢測能力，從而提高了模型在實時應用中的準確性和穩定性。

首先，我們設計了一種簡單又有效的小物件注意力模組SWS(SimAM With Slicing)，並在PRB-FPN的基礎上整合了CBAM和DSConv，進一步提升模型性能，改進後的模型稱為DCT-PRB。在VisDrone2019-testset-dev數據集上的大量實驗表明，DCT-PRB在各項指標上均優於目前主流的單階段檢測模型，比YOLOv10x、RT-DETR-X、YOLOv7x、PRB-FPN分別高出5.7%、8.3%、1.2%和0.7%的mAP，且FPS保持在101。在注意力模組的對比實驗中，SWS也達到了最佳效能，超越了CBAM、SE、SimAM等主流注意力模組，其參數量只占評分第二名Biformer的10.15%，顯示出SWS模組在小物件辨識上的強大實力。

此外，我們在DCT-PRB的基礎上結合了物件追蹤模型SMILEtrack，通過設計多項約束和動態衝突追蹤框，實現了可應用於串流環境的潛在人車衝突檢測。相較於先前的事故點檢測研究，我們的檢測策略不再局限於特定地區，也不受特定情況的限制，顯示出其廣泛的應用潛力。

Drones provide a comprehensive perspective for observing the environment, significantly reducing object deformation and occlusion issues, and making them advantageous for monitoring traffic and safety conditions. In this study, we propose a dynamic pedestrian-vehicle collision detection model based on drone perspectives for quickly and accurately detecting potential pedestrian risks on roads. To enhance detection performance, we have made improvements to the model, particularly augmenting its capability to detect small objects, thereby increasing the accuracy and stability of the model in real-time applications.

Firstly, we designed a simple yet effective small object attention module called SWS (SimAM With Slicing) and integrated CBAM and DSConv into the PRB-FPN foundation to enhance model performance further. The improved model is named DCT-PRB. Extensive experiments on the VisDrone2019-test set-dev dataset show that DCT-PRB outperforms current mainstream single-stage detection models, exceeding YOLOv10x, RT-DETR-X, YOLOv7x, and PRB-FPN by 5.7%, 8.3%, 1.2%, and 0.7% in mAP, respectively, while maintaining an FPS of 101. In the attention module comparison experiments, SWS also achieved the best performance, surpassing mainstream attention modules such as CBAM, SE, and SimAM, with its parameter count being only 10.15% of the second-ranked Biformer, demonstrating the strong capability of the SWS module in small object recognition.

Additionally, we combined the object tracking model SMILEtrack with DCT-PRB, designing multiple constraints and dynamic collision tracking frames to achieve potential pedestrian-vehicle collision detection applicable in streaming environments. Compared to previous accident point detection studies, our detection strategy is no longer confined to specific regions or situations, showcasing its broad application potential.

摘要 I
Abstract II
目次 III
表目錄 V
圖目錄 VI
第一章、緒論 1
1.1研究背景與動機 1
1.2研究問題與重要性 3
1.3研究貢獻 4
1.4論文架構 4
第二章、文獻回顧 5
2.1物件辨識任務的演進和種類 5
2.1.1卷積神經網路的演進 5
2.1.2雙階段目標檢測 7
2.1.3單階段目標檢測 9
2.2卷積結構的改進 11
2.3單階段物件辨識模型的改進 13
2.4圖像注意力機制的改進 15
2.5多目標追蹤 17
2.6 智慧交通系統 18
2.6.1基於感測數據與模擬 18
2.6.2基於平面道路影像 19
2.6.3基於無人機影像 20
第三章、研究方法 23
3.1數據預處理 23
3.1.1取得資料集 23
3.1.2 標註訊息處理 24
3.1.3 訓練資料隨機減半 27
3.1.4 數據可視化 28
3.2建立檢測模型 29
3.2.1 DCT-PRB網路結構 29
3.2.2 無參數注意力模組SWS 29
3.2.3 ELAN-S模組 31
3.2.4 ELAN-D模組 32
3.2.5 S2PC-CBAM模組 33
3.3 道路人車衝突偵測模型 34
3.3.1 物件檢測與匹配系統 36
3.3.2 互動式車道輪廓繪制介面 36
3.3.3 衝突檢測規則 37
3.3.4 生成潛在衝突熱點圖 38
第四章、實驗結果與討論 39
4.1實驗環境與實驗參數 39
4.2評估指標 40
4.3模型對比實驗 41
4.4注意力模組對比實驗 43
4.5消融性實驗 44
4.6動態人-車潛在衝突檢測模型成果 45
4.6.1互動式道路輪廓繪製介面效果 45
4.6.2追蹤模型效果 47
4.6.3約束效果 48
4.6.4潛在衝突熱點圖-案例分析 49
第五章、結論與未來展望 53
特別感謝 54
參考文獻 55

[1]D. Qiu, Q. Xu, and J. Zhang, "Improvement of pedestrian crossing safety on urban roads," in 2010 International Conference on Intelligent Computation Technology and Automation, 2010, vol. 2: IEEE, pp. 514-517.
[2]P. Tumas, A. Nowosielski, and A. Serackis, "Pedestrian detection in severe weather conditions," IEEE Access, vol. 8, pp. 62775-62784, 2020.
[3]A. Narayanan, R. D. Kumar, R. RoselinKiruba, and T. S. Sharmila, "Study and analysis of pedestrian detection in thermal images using YOLO and SVM," in 2021 Sixth International Conference on Wireless Communications, Signal Processing and Networking (WiSPNET), 2021: IEEE, pp. 431-434.
[4]Y.-T. Huang and L. Tzu-Chang, "A binomial distribution model for describing pedestrian-vehicle crashes in urban areas," Asian Transport Studies, vol. 10, p. 100131, 2024.
[5]S. Roy and M. S. Rahman, "Emergency vehicle detection on heavy traffic road from cctv footage using deep convolutional neural network," in 2019 international conference on electrical, computer and communication engineering (ECCE), 2019: IEEE, pp. 1-6.
[6]H. Xiao et al., "An improved LSTM model for behavior recognition of intelligent vehicles," IEEE Access, vol. 8, pp. 101514-101527, 2020.
[7]H. Zheng, F. Lin, X. Feng, and Y. Chen, "A hybrid deep learning model with attention-based conv-LSTM networks for short-term traffic flow prediction," IEEE Transactions on Intelligent Transportation Systems, vol. 22, no. 11, pp. 6910-6920, 2020.
[8]K. D. Dharmadasa, G. K. Sahoo, S. K. Das, and P. Singh, "Video-based road accident detection on highways: A less complex YOLOv5 approach," in 2023 International Conference on Computer, Electronics & Electrical Engineering & their Applications (IC2E3), 2023: IEEE, pp. 1-6.
[9]A. Bhadange, V. Jabade, R. Bhattad, and P. Daga, "A Smart Surveillance System for Video-Based Vehicle Detection and Speed Monitoring," in 2024 2nd International Conference on Advancement in Computation & Computer Technologies (InCACCT), 2024: IEEE, pp. 689-694.
[10]J. Lin, Q. Luo, Y. Qian, X. Zhang, C. Wang, and M. Yang, "A Vision-based Pedestrian Comity Pre-Warning System," in 2021 International Conference on Security, Pattern Analysis, and Cybernetics (SPAC), 2021: IEEE, pp. 385-391.
[11]S. I. S. Alquijay and R. S. Wang, "Smart City Traffic Lights Management System: A Case Study to Improve Crosswalk Safety in Taipei," in 2019 IEEE International Conference on Consumer Electronics-Taiwan (ICCE-TW), 2019: IEEE, pp. 1-2.
[12]R. Quan, L. Zhu, Y. Wu, and Y. Yang, "Holistic LSTM for pedestrian trajectory prediction," IEEE transactions on image processing, vol. 30, pp. 3229-3239, 2021.
[13]K. Hara, H. Kataoka, M. Inaba, K. Narioka, R. Hotta, and Y. Satoh, "Predicting appearance of vehicles from blind spots based on pedestrian behaviors at crossroads," IEEE transactions on intelligent transportation systems, vol. 23, no. 8, pp. 11917-11929, 2021.
[14]J. Chen, J. Zhu, R. Xu, Y. Chen, H. Zeng, and J. Huang, "ORNet: Orthogonal Re-parameterized Networks for Fast Pedestrian and Vehicle Detection," IEEE Transactions on Intelligent Vehicles, 2023.
[15]J. Lee and K. Jang, "Proactive detection of crash hotspots using in-vehicle driving recorder," in 2016 3rd Asia-Pacific World Congress on Computer Science and Engineering (APWC on CSE), 2016: IEEE, pp. 193-198.
[16]T. Suzuki, K. Yamori, C. Zhang, T. Miyoshi, and Y. Tanaka, "Relation between Warning Error and Vehicle Speed in Vehicle-to-Pedestrian Warning System," in 2021 22nd Asia-Pacific Network Operations and Management Symposium (APNOMS), 2021: IEEE, pp. 29-32.
[17]S. Zhankaziev, A. Zamytskih, A. Vorobyev, M. Gavrilyuk, and M. Pletnev, "Predicting traffic accidents using the conflict coefficient," in 2022 Intelligent Technologies and Electronic Devices in Vehicle and Road Transport Complex (TIRVED), 2022: IEEE, pp. 1-6.
[18]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 Proceedings of the IEEE/CVF international conference on computer vision, 2021, pp. 2778-2788.
[19]Y.-H. Chung and C.-Y. Su, "Object Detection Algorithm Based on Improved YOLOv7 for UAV Images," in 2023 IEEE 5th Eurasia Conference on IOT, Communication and Engineering (ECICE), 2023: IEEE, pp. 18-21.
[20]X. Li, R. Zhu, X. Yu, Y. Cong, and G. Liu, "Symmetric Joint Detection and Re-Identification for UAV-Based Multiple Object Tracking," in 2023 42nd Chinese Control Conference (CCC), 2023: IEEE, pp. 8532-8538.
[21]A. Chen, D. Zheng, H. Yang, D. Li, R. Hou, and H. Liu, "Design of Multi-Sensor Array Payload for UAV Bridge Concrete Surface Crack Detection," in 2023 IEEE 16th International Conference on Electronic Measurement & Instruments (ICEMI), 2023: IEEE, pp. 24-29.
[22]Z. Cao, L. Kooistra, W. Wang, L. Guo, and J. Valente, "Real-time object detection based on uav remote sensing: A systematic literature review," Drones, vol. 7, no. 10, p. 620, 2023.
[23]B.-S. Wang et al., "PRB-FPN+: Video analytics for enforcing motorcycle helmet laws," in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2023, pp. 5476-5484.
[24]Y. Wang, I. W.-H. Ho, and Y. Wang, "Real-time intersection vehicle turning movement counts from live UAV video stream using multiple object tracking," Journal of Intelligent and Connected Vehicles, 2023.
[25]M. A. Khan, W. Ectors, T. Bellemans, D. Janssens, and G. Wets, "Unmanned aerial vehicle-based traffic analysis: A case study for shockwave identification and flow parameters estimation at signalized intersections," Remote Sensing, vol. 10, no. 3, p. 458, 2018.
[26]P.-Y. Chen et al., "Drone-Based Vehicle Flow Estimation and its Application to Traffic Conflict Hotspot Detection at Intersections," in 2020 IEEE International Conference on Image Processing (ICIP), 2020: IEEE, pp. 1521-1525.
[27]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.
[28]Y. Zhao et al., "Detrs beat yolos on real-time object detection," arXiv preprint arXiv:2304.08069, 2023.
[29]P.-Y. Chen, M.-C. Chang, J.-W. Hsieh, and Y.-S. Chen, "Parallel residual bi-fusion feature pyramid network for accurate single-shot object detection," IEEE Transactions on Image Processing, vol. 30, pp. 9099-9111, 2021.
[30]C. Hao, H. Zhang, W. Song, F. Liu, and E. Wu, "SliNet: Slicing-Aided Learning for Small Object Detection," IEEE Signal Processing Letters, 2024.
[31]F. Yu and V. Koltun, "Multi-scale context aggregation by dilated convolutions," arXiv preprint arXiv:1511.07122, 2015.
[32]J. Dai et al., "Deformable convolutional networks," in Proceedings of the IEEE international conference on computer vision, 2017, pp. 764-773.
[33]Y. Qi, Y. He, X. Qi, Y. Zhang, and G. Yang, "Dynamic snake convolution based on topological geometric constraints for tubular structure segmentation," in Proceedings of the IEEE/CVF International Conference on Computer Vision, 2023, pp. 6070-6079.
[34]J. Hu, L. Shen, and G. Sun, "Squeeze-and-excitation networks," in Proceedings of the IEEE conference on computer vision and pattern recognition, 2018, pp. 7132-7141.
[35]S. Woo, J. Park, J.-Y. Lee, and I. S. Kweon, "Cbam: Convolutional block attention module," in Proceedings of the European conference on computer vision (ECCV), 2018, pp. 3-19.
[36]Q. Hou, D. Zhou, and J. Feng, "Coordinate attention for efficient mobile network design," in Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, 2021, pp. 13713-13722.
[37]L. Yang, R.-Y. Zhang, L. Li, and X. Xie, "Simam: A simple, parameter-free attention module for convolutional neural networks," in International conference on machine learning, 2021: PMLR, pp. 11863-11874.
[38]L. Zhu, X. Wang, Z. Ke, W. Zhang, and R. W. Lau, "Biformer: Vision transformer with bi-level routing attention," in Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, 2023, pp. 10323-10333.
[39]H. Rezatofighi, N. Tsoi, J. Gwak, A. Sadeghian, I. Reid, and S. Savarese, "Generalized intersection over union: A metric and a loss for bounding box regression," in Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, 2019, pp. 658-666.
[40]Z. Zheng, P. Wang, W. Liu, J. Li, R. Ye, and D. Ren, "Distance-IoU loss: Faster and better learning for bounding box regression," in Proceedings of the AAAI conference on artificial intelligence, 2020, vol. 34, no. 07, pp. 12993-13000.
[41]Y.-F. Zhang, W. Ren, Z. Zhang, Z. Jia, L. Wang, and T. Tan, "Focal and efficient IOU loss for accurate bounding box regression," Neurocomputing, vol. 506, pp. 146-157, 2022.
[42]Z. Tong, Y. Chen, Z. Xu, and R. Yu, "Wise-IoU: Bounding Box Regression Loss with Dynamic Focusing Mechanism," arXiv preprint arXiv:2301.10051, 2023.
[43]R. Girshick, J. Donahue, T. Darrell, and J. Malik, "Rich feature hierarchies for accurate object detection and semantic segmentation," in Proceedings of the IEEE conference on computer vision and pattern recognition, 2014, pp. 580-587.
[44]R. Girshick, "Fast r-cnn," in Proceedings of the IEEE international conference on computer vision, 2015, pp. 1440-1448.
[45]S. Ren, K. He, R. Girshick, and J. Sun, "Faster r-cnn: Towards real-time object detection with region proposal networks," Advances in neural information processing systems, vol. 28, 2015.
[46]J. Redmon, S. Divvala, R. Girshick, and A. Farhadi, "You only look once: Unified, real-time object detection," in Proceedings of the IEEE conference on computer vision and pattern recognition, 2016, pp. 779-788.
[47]W. Liu et al., "Ssd: Single shot multibox detector," in Computer Vision–ECCV 2016: 14th European Conference, Amsterdam, The Netherlands, October 11–14, 2016, Proceedings, Part I 14, 2016: Springer, pp. 21-37.
[48]S. Liu and D. Huang, "Receptive field block net for accurate and fast object detection," in Proceedings of the European conference on computer vision (ECCV), 2018, pp. 385-400.
[49]X. Zhu, H. Hu, S. Lin, and J. Dai, "Deformable convnets v2: More deformable, better results," in Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, 2019, pp. 9308-9316.
[50]A. Wang et al., "YOLOv10: Real-Time End-to-End Object Detection," arXiv preprint arXiv:2405.14458, 2024.
[51]Y.-H. Wang, "SMILEtrack: SiMIlarity LEarning for Multiple Object Tracking," arXiv preprint arXiv:2211.08824, 2022.
[52]Y. Zhang et al., "Bytetrack: Multi-object tracking by associating every detection box," in European conference on computer vision, 2022: Springer, pp. 1-21.
[53]N. Aharon, R. Orfaig, and B.-Z. Bobrovsky, "BoT-SORT: Robust associations multi-pedestrian tracking," arXiv preprint arXiv:2206.14651, 2022.
[54]N. Sharma and R. D. Garg, "Real-time iot-based connected vehicle infrastructure for intelligent transportation safety," IEEE Transactions on Intelligent Transportation Systems, 2023.
[55]Ö. Aköz and M. E. Karslıgil, "Video-based traffic accident analysis at intersections using partial vehicle trajectories," in 2010 IEEE 18th Signal Processing and Communications Applications Conference, 2010: IEEE, pp. 499-502.
[56]Q. Zhao, B. Liu, S. Lyu, C. Wang, and H. Zhang, "TPH-YOLOv5++: Boosting Object Detection on Drone-Captured Scenarios with Cross-Layer Asymmetric Transformer," Remote Sensing, vol. 15, no. 6, p. 1687, 2023.
[57]A. Panthakkan, W. Mansoor, and H. Al Ahmad, "Accurate UAV-Based Vehicle Detection: The Cutting-Edge YOLOv7 Approach," in 2023 International Symposium on Image and Signal Processing and Analysis (ISPA), 2023: IEEE, pp. 1-4.
[58]M. Muzammul, A. M. Algarni, Y. Y. Ghadi, and M. Assam, "Enhancing UAV Aerial Image Analysis: Integrating Advanced SAHI Techniques with Real-Time Detection Models on the VisDrone Dataset," IEEE Access, 2024.
[59]Z. Ge, L. Qi, Y. Wang, and Y. Sun, "Zoom-and-reasoning: Joint foreground zoom and visual-semantic reasoning detection network for aerial images," IEEE Signal Processing Letters, vol. 29, pp. 2572-2576, 2022.
[60]S. K. Gonuguntla, "A topological analysis of pedestrian crashes in Citta'Studi, Milano," 2022.