建築空間的安全管理因為其牽涉到人與空間交互關係的複雜性，一向極度仰賴人力監控。根據過往統計，許多既成建築場域中的事故常因人員未察覺潛在風險環境而發生，如半開放車道出入口、施工修繕區或不易察覺的磁磚剝落區以及設備吊掛區。現有安全監控多屬事後調閱，缺乏即時預測與預警能力。為提升預防性危安管理效能，如何即時掌握行人動向、預測其移動軌跡，並依其與危險區域的相對位置評估風險程度，是發展智慧安全系統之重要方向。
本研究旨在建立一套可即時辨識與預測行人軌跡之系統，透過深度學習與時間序列模型，預測行人未來位置，並依據預測點與危險區的空間關係進行風險計算，進而提供行人與管理單位即時預警，降低潛在危險發生機率。
整體系統包含三大功能模組：一為行人辨識與ID追蹤，使用監視影像進行目標標定與持續追蹤；二為軌跡預測模型，導入 Kalman Filter、多項式回歸與LSTM三種方法進行比較與驗證；三為風險值計算與視覺化呈現，透過距離公式轉換為風險百分比，實現可視化警示系統。
本研究系統不僅可提升監控效率與空間理解能力，也為建築物業管理、自動化控制與智慧安全系統提供一項基礎架構與應用潛力，助益於未來智慧場域管理之實務提供具即時性與預防性的輔助工具。

Safety management in architectural spaces has long relied heavily on manual supervision due to the inherent complexity of human–space interactions. Traditional surveillance systems primarily rely on post-incident video review and lack real-time understanding of human behavior. Many accidents occur due to individuals failing to detect potential hazards such as open access driveways, construction zones, or damaged flooring.
This study proposes an integrated framework for real-time pedestrian recognition, trajectory prediction, and risk assessment, with the aim of strengthening proactive safety management in architectural environments. By combining computer vision techniques with time-series forecasting models, the system seeks not only to detect pedestrians but also to anticipate their future movement patterns and evaluate potential risks. Such predictive capabilities offer new possibilities for reducing accident probability, guiding crowd movement, and supporting intelligent facility management. The design of this framework is informed by three primary modules: pedestrian identification and ID-based tracking, trajectory prediction using multiple algorithms, and risk visualization through path overlays.

[1] admin. (2018, September 12). OpenCV3 Python之MOG2、KNN和GMG背景分割器 [Background subtractors MOG2, KNN, and GMG in OpenCV3 Python]. Retrieved from https://www.yyearth.com/index.php?aid=243
[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] 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. https://doi.org/10.1109/ICIP.2016.7533003
[4] Bommasani, R., Hudson, D. A., Adeli, E., Altman, R., Arora, S., von Arx, S., ... & Liang, P. (2021). On the opportunities and risks of foundation models. arXiv preprint arXiv:2108.07258.
[5] Braham, M., & Van Droogenbroeck, M. (2017). Semantic background subtraction. International Journal of Computer Vision, 123(3), 601–622.
[6] Braham, M., Piérard, S., & Van Droogenbroeck, M. (2017, September). Semantic background subtraction. In 2017 IEEE International Conference on Image Processing (ICIP) (pp. 4552-4556). Ieee.
[7] Cheng, H., Liu, M., & Chen, L. (2023, September). An end-to-end framework of road user detection, tracking, and prediction from monocular images. In 2023 IEEE 26th International Conference on Intelligent Transportation Systems (ITSC) (pp. 2178-2185). IEEE.
[8] Dalal, N., & Triggs, B. (2005, June). Histograms of oriented gradients for human detection. In 2005 IEEE computer society conference on computer vision and pattern recognition (CVPR'05) (Vol. 1, pp. 886-893). Ieee.https://doi.org/10.1109/CVPR.2005.177
[9] Gaidon, A., Wang, Q., Cabon, Y., & Vig, E. (2016). Virtual worlds as proxy for multi-object tracking analysis. In Proceedings of the IEEE conference on computer vision and pattern recognition (pp. 4340-4349). https://doi.org/10.1109/CVPR.2016.470
[10] Giuliari, F., Hasan, I., Cristani, M., & Galasso, F. (2021, January). Transformer networks for trajectory forecasting. In 2020 25th international conference on pattern recognition (ICPR) (pp. 10335-10342). IEEE.
[11] Gupta, A., Johnson, J., Fei-Fei, L., Savarese, S., & Alahi, A. (2018). Social gan: Socially acceptable trajectories with generative adversarial networks. In Proceedings of the IEEE conference on computer vision and pattern recognition (pp. 2255-2264).
[12] Helbing, D., & Molnar, P. (1995). Social force model for pedestrian dynamics. Physical review E, 51(5), 4282.
[13] Helbing, D., & Molnar, P. (1995). Social force model for pedestrian dynamics. Physical review E, 51(5), 4282.https://doi.org/10.1103/PhysRevE.51.4282
[14] Hsu, B. (2023, April 19). [圖像處理] 二值化閥值自動化篩選 – Otsu, 多重門檻值, 直方圖 https://medium.com/@benhsu501/二值化閥值自動化篩選-otsu-多重門檻值-直方圖-345aff032e0f
[15] Jalal, A. S., & Singh, V. (2014). A framework for background modelling and shadow suppression for moving object detection in complex wavelet domain. Multimedia tools and applications, 73, 779-801.
[16] Jalal, A. S., & Singh, V. (2014). A framework for background modelling and shadow suppression for moving object detection in complex wavelet domain. Multimedia Tools and Applications, 72(2), 221–245.
[17] Kadir, K., Kamaruddin, M. K., Nasir, H., Safie, S. I., & Bakti, Z. A. K. (2014, August). A comparative study between LBP and Haar-like features for Face Detection using OpenCV. In 2014 4th International conference on engineering technology and technopreneuship (ICE2T) (pp. 335-339). IEEE.https://doi.org/10.1109/ICE2T.2014.7006273
[18] Lienhart, R., & Maydt, J. (2002, September). An extended set of haar-like features for rapid object detection. In Proceedings. international conference on image processing (Vol. 1, pp. I-I). IEEE. https://doi.org/10.1109/ICIP.2002.1038171
[19] Mallick, S. (2016, November 14). Image recognition and object detection – Part 1 [Blog post]. LearnOpenCV. https://learnopencv.com/image-recognition-and-object-detection-part1/
[20] OpenCV. (n.d.). Planar object detection using homography [Illustration]. OpenCV Documentation. Retrieved June 18, 2025, from https://docs.opencv.org/4.x/d9/dab/tutorial_homography.html
[21] Park, S., & Trivedi, M. M. (2007, February). Homography-based analysis of people and vehicle activities in crowded scenes. In 2007 IEEE Workshop on Applications of Computer Vision (WACV'07) (pp. 51-51). IEEE.https://doi.org/10.1109/WACV.2007.42
[22] Parquery AG. (2022). Control adaptive traffic signals with cameras, not induction loops [Image]. Parquery. Retrieved from https://parquery.com/control-traffic-signals-with-cameras-not-induction-loops/
[23] Rhodes, A. (n.d.). Computer Vision: Multiple Object Tracking (MOT). Portland State University. Retrieved June 18, 2025, from https://web.pdx.edu/~arhodes/CV_MOT.pdf
[24] Shahriar, K. N. H. (2023, February 1). What is convolutional neural network — CNN (Deep Learning). Medium. Retrieved from https://nafizshahriar.medium.com/what-is-convolutional-neural-network-cnn-deep-learning-b3921bdd82d5
[25] SharkYun. (2024, October 31). Object detection, one stage vs two stage detectors. Medium. https://sharkyun.medium.com/computer-vision-object-detection-one-stage-vs-two-stage-b05dbff88195
[26] Song, L., Wu, J., Yang, M., Zhang, Q., Li, Y., & Yuan, J. (2021). Stacked homography transformations for multi-view pedestrian detection. In Proceedings of the IEEE/CVF International Conference on Computer Vision (pp. 6049-6057).
[27] Teicholz, P. (Ed.). (2013). BIM for facility managers. John Wiley & Sons..
[28] Trnovszký, T., Sýkora, P., & Hudec, R. (2017). Comparison of background subtraction methods on near infra-red spectrum video sequences. Procedia engineering, 192, 887-892. https://doi.org/10.1016/j.proeng.2017.06.153
[29] Tryolabs. (2023). Norfair documentation: High performance, lightweight library for multi-object tracking. Retrieved June 18, 2025, from https://tryolabs.github.io/norfair/
[30] Wavelet. (n.d.). In Wikipedia. Retrieved July 5, 2025, from https://en.wikipedia.org/wiki/Wavelet
[31] Welch, G., & Bishop, G. (1995). An introduction to the Kalman filter. from https://www.cs.unc.edu/~welch/media/pdf/kalman_intro.pdf
[32] Wojke, N., Bewley, A., & Paulus, D. (2017, September). Simple online and realtime tracking with a deep association metric. In 2017 IEEE international conference on image processing (ICIP) (pp. 3645-3649). IEEE.
[33] Zhang, Z. (2002). A flexible new technique for camera calibration. IEEE Transactions on pattern analysis and machine intelligence, 22(11), 1330-1334.https://doi.org/10.1109/34.888718
[34] Zivkovic, Z. (2004, August). Improved adaptive Gaussian mixture model for background subtraction. In Proceedings of the 17th International Conference on Pattern Recognition, 2004. ICPR 2004. (Vol. 2, pp. 28-31). IEEE.
[35] Zou, Z., Chen, K., Shi, Z., Guo, Y., & Ye, J. (2023). Object detection in 20 years: A survey. Proceedings of the IEEE, 111(3), 257-276.https://doi.org/10.48550/arXiv.1905.05055
[36] 王榮進、沈揚庭（2020）。應用人工智慧科技提升建築物維運管理效益之研究。內政部建築研究所協同研究。
[37] 林瑞浤（2024）。主動式BIM:結合電腦視覺與數位雙生的場域人流管理系統〔碩士論文，成功大學〕。
[38] 黃亭維（2024）。群眾行為模式分析模型-基於機器學習的分群學習法〔碩士論文，成功大學〕。
[39] 廖士豪（2020）。整合AI電腦視覺與BIM電子圍籬發展智慧維運平台〔碩士論文，逢甲大學〕。