目前新北市、台北市、桃園市、高雄市在道路挖掘施工時皆規定需要在 工地架設攝影機來即時錄影監控，本研究為了能監控攝影模組的狀態，透過 SSTP 與攝影模組的路由器建立連線以取得溫度、電壓等數值，再使用 Node.js 建立監控平台，記錄攝影機的運作狀態，並偵測回傳的數值，若發現異常狀 態，則使用 Line Notify 推播，以降低監控人員需要觀看螢幕的時間，並更容 易找出設備異常的可能原因。
此外，本研究運用 YOLOv5 深度學習之方式建立職安狀態辨識模型， 並與其他物件偵測演算法比較。使用模型即時對施工監控影像進行物件偵測， 記錄違規的樣態，如未配戴安全帽或未配戴反光背心的施工人員，系統將違 規的時間點記錄下來，若超出一定的時間範圍就以 Line Notify 推播，期望減 少施工時發生意外的可能性。

At present, New Taipei City, Taipei City, Taoyuan City, and Kaohsiung City stipulate cameras need to be erected on the construction site for real-time video monitoring during road excavation construction. To monitor the status of the camera module, this study establishes a connection with the camera module router through SSTP. After obtaining temperature, voltage, and signal strength values, use Node.js to build a monitoring platform, record the operating status of the camera, and detect the returned values. If the value is abnormal, the system will send warning messages. Those reduce the time of checking the status and make identifying possible causes of device abnormalities easier.
In addition, this study uses the YOLOv5, a deep learning method, to establish an occupational safety status identification model and compares it with other object detection algorithms. Use the model to detect objects in construction monitor images promptly and save the violation record to the database. For example, construction workers who do not wear safety helmets or reflective vests. If they exceed a specific time range, the system will use Line Notify to send the warning messages, hoping to reduce the possibility of accidents during construction.

[1]  L. Wang et al., “Automatic Monitoring System in Underground Engineering Construction: Review and Prospect,” Advances in Civil Engineering, vol. 2020, p. 3697253, Jun. 2020.
[2]  N.DalalandB.Triggs,“Histogramsoforientedgradientsforhumandetection,” in 2005 IEEE Computer Society Conference on Computer Vision and Pattern Recognition (CVPR’05), vol. 1, pp. 886–893, Jun. 2005.
[3]  D. G. Lowe, “Object recognition from local scale-invariant features,” in Proceedings of the Seventh IEEE International Conference on Computer Vision, vol. 2, pp. 1150–1157, Sep. 1999.
[4]  PyLessons, YOLOv3 theory explained, [Online], https://pylessons.com/YOLOv3-introduction
[5]  J. Redmon, S. Divvala, R. Girshick, and A. Farhadi, “You Only Look Once: Unified, Real-Time Object Detection,” arXiv, arXiv:1506.02640, May 2016.
[6]  W. Liu et al., “SSD: Single Shot MultiBox Detector,” in Computer Vision – ECCV 2016, pp. 21–37, 2016.
[7]  R. Girshick, J. Donahue, T. Darrell, and J. Malik, “Rich Feature Hierarchies for Accurate Object Detection and Semantic Segmentation,” pp. 580–587, 2014.
[8]  R. Girshick, “Fast R-CNN,” in 2015 IEEE International Conference on Computer Vision (ICCV), pp. 1440–1448, Dec. 2015.
[9]  S. Ren, K. He, R. Girshick, and J. Sun, “Faster R-CNN: Towards Real-Time Object Detection with Region Proposal Networks,” in Advances in Neural Information Processing Systems, vol. 28, 2015.
[10] Y. Li, H. Wei, Z. Han, J. Huang, and W. Wang, “Deep Learning-Based Safety Helmet Detection in Engineering Management Based on Convolutional Neural Networks,” Advances in Civil Engineering, vol. 2020, p. 9703560, Sep. 2020.
[11] Z. Jin et al., “DWCA-YOLOv5: An Improve Single Shot Detector for Safety Helmet Detection,” Journal of Sensors, vol. 2021, p. 4746516, Oct. 2021. [12] Glenn Jocher. Yolov5. Code repository https://github.com/ultralytics/yolov5. [13]行政院公報第 026 卷第 112 期衛生勞動, 2020, 檢自
https://gazette.nat.gov.tw/EG_FileManager/eguploadpub/eg026112/ch08/typ
e3/gov82/num34/Eg.pdf.
[14] Make Sense, [Online], https://www.makesense.ai.
[15] Shinobi, [Online], https://shinobi.video.
[16] TimescaleDB, [Online], https://www.timescale.com.
[17] Mike Freedman, Time-series data: Why (and how) to use a relational database
instead of NoSQL, 2017, [Online], Available: https://www.timescale.com/blog/time-series-data-why-and-how-to-use-a- relational-database-instead-of-nosql-d0cd6975e87c/
[18] M. Martinviita, “Time series database in Industrial IoT and its testing tool,” University of Oulu, Master’s Thesis, 2018.
[19] Grafana, [Online], https://grafana.com.
[20] J. Redmon and A. Farhadi, “YOLOv3: An Incremental Improvement,” arXiv,
arXiv:1804.02767, Apr. 2018.
[21] Z. Zheng, P. Wang, W. Liu, J. Li, R. Ye, and D. Ren, “Distance-IoU Loss:
Faster and Better Learning for Bounding Box Regression,” Proceedings of the AAAI Conference on Artificial Intelligence, vol. 34, no. 07, Art. no. 07, Apr. 2020.
[22] PyTorch, Binary Cross Entropy Loss, [Online], https://pytorch.org/docs/stable/generated/torch.nn.BCELoss.html
[23]王 定 邦 , Anchor Boxes 示 意 圖 , [Online], https://blog.csdn.net/weixin_44238733/article/details/121180095?spm=1001. 2101.3001.6650.3&utm_medium=distribute.pc_relevant.none-task-blog- 2%7Edefault%7ECTRLIST%7ERate-3.pc_relevant_default&depth_1- utm_source=distribute.pc_relevant.none-task-blog- 2%7Edefault%7ECTRLIST%7ERate- 3.pc_relevant_default&utm_relevant_index=6
[24] scikit-learn, Precision-Recall 曲 線 示 意 圖 , [Online], https://scikit- learn.org/stable/auto_examples/model_selection/plot_precision_recall.html
[25] D. M. W. Powers, “Evaluation: from precision, recall and F-measure to ROC, informedness, markedness and correlation,” arXiv, arXiv:2010.16061, Oct. 2020.
[26] J. D. Rodriguez, A. Perez, and J. A. Lozano, “Sensitivity Analysis of k-Fold Cross Validation in Prediction Error Estimation,” IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 32, no. 3, pp. 569–575, Mar. 2010.
[27] K-Fold Cross-Validation, [Online], https://androidkt.com/pytorch-k-fold- cross-validation-using-dataloader-and-sklearn/
[28] A. Bochkovskiy, C.Y. Wang, and H.Y. M. Liao, “YOLOv4: Optimal Speed and Accuracy of Object Detection,” arXiv, arXiv:2004.10934, Apr. 2020.
[29] C.Y. Wang, I.H. Yeh, and H.-Y. M. Liao, “You Only Learn One Representation: Unified Network for Multiple Tasks,” arXiv, arXiv:2105.04206, May 2021.