系統識別號 | U0002-0909202318410400 |
---|---|
DOI | 10.6846/tku202300652 |
論文名稱(中文) | 在居家環境中以深度學習預防老人跌倒之危險物偵測技術 |
論文名稱(英文) | Using deep learning for detecting hazardous objects to prevent falls in a home environment |
第三語言論文名稱 | |
校院名稱 | 淡江大學 |
系所名稱(中文) | 資訊工程學系碩士班 |
系所名稱(英文) | Department of Computer Science and Information Engineering |
外國學位學校名稱 | |
外國學位學院名稱 | |
外國學位研究所名稱 | |
學年度 | 111 |
學期 | 2 |
出版年 | 112 |
研究生(中文) | 黃厚鈞 |
研究生(英文) | HOU-CHUN HUANG |
學號 | 610410689 |
學位類別 | 碩士 |
語言別 | 繁體中文 |
第二語言別 | |
口試日期 | 2023-06-12 |
論文頁數 | 45頁 |
口試委員 |
指導教授
-
張志勇(cychang@mail.tku.edu.tw)
共同指導教授 - 郭經華(chkuo@mail.tku.edu.tw) 口試委員 - 廖文華 口試委員 - 林怡第 |
關鍵字(中) |
深度學習 人工智慧 電腦視覺 影像辨識 深度影像圖 孿生網路 YOLO ChatGPT |
關鍵字(英) |
Deep Learning Artificial Intelligence Computer Vision Image Recognition Depth Image YOLO Siamese Neural Network ChatGPT |
第三語言關鍵字 | |
學科別分類 | |
中文摘要 |
近年來,影像辨識,作為人工智慧領域中的一個重要分支,已經在過去幾年中取得了巨大的進步和突破。影像辨識技術對於各個領域都有著廣泛的應用,從醫學診斷到自動駕駛,皆因其能夠幫助我們從大量的視覺數據中提取有用的訊息。而影像辨識技術在預防老人跌倒方面扮也演了關鍵角色,為日益增長的老年人口提供了更好的照護和安全保障。這項技術利用先進的影像辨識方法,通過監測老人的行動和環境,早期識別潛在的跌倒風險,並採取適當的措施來預防意外事故。然而,大多數的辨識技術僅以辨識老人的肢體動作為是否跌倒的依據,並不能及時預防老人跌倒。 本篇論文旨在利用老人居家的即時影像,透過將影像轉深度圖,並使用孿生神經網路,比對是否出現不應該存在的障礙物。而這些障礙物將被分類為不同的危險程度,以幫助評估其威脅程度。一旦識別到潛在危險的障礙物,系統將立即透過通訊軟體中的ChatGPT通知老人。 ChatGPT將提供即時的通知,提醒老人有障礙物存在,同時給予指導,協助他們採取迅速的措施來避免潛在的風險,確保老人的生命安全。這項研究旨在提供一個智能且敏捷的方式,以維護老年人的居家安全和福祉。 |
英文摘要 |
In recent years, image recognition, as a critical branch within the domain of artificial intelligence, has seen significant advancements and breakthroughs. Image recognition technology has widespread applications across various sectors, from medical diagnoses to autonomous driving, due to its ability to extract meaningful information from vast visual data. Image recognition also plays a pivotal role in elderly fall prevention, providing enhanced care and security for the growing elderly population. This technology employs sophisticated image recognition techniques to monitor elderly activities and their environment, identifying potential fall risks early and taking appropriate measures to prevent accidents. However, most recognition technologies merely use the elderly's limb movements as the basis for fall detection, failing to prevent falls in real-time. This paper aims to utilize real-time home imaging of the elderly, converting the images to depth maps, and employing Siamese neural networks to detect the presence of unintended obstacles. These obstacles are categorized based on their risk levels to help assess their potential threat. Once a potentially dangerous obstacle is identified, the system immediately notifies the elderly through the ChatGPT function in communication software. ChatGPT offers real-time alerts, warning the elderly of the obstacle's presence and providing guidance to help them take swift measures to mitigate potential risks, ensuring their safety. This research seeks to provide an intelligent and agile means to uphold the safety and well-being of the elderly at home. |
第三語言摘要 | |
論文目次 |
目錄 目錄 VII 圖目錄 IX 表目錄 X 第一章 緒論 1 1-1研究背景 1 1-2研究動機 2 1-3研究目的 3 1-4研究貢獻 3 第二章 文獻探討 6 2-1 傳統骨架辨識技術 6 2-2基於YOLO辨識技術 9 第三章 前景知識 15 3-1 YOLO辨識居家物品 15 3-1-1 YOLO 15 3-2 居家影像深度處理 18 3-2-1 深度影像 18 3-3 孿生神經網路對比家中多餘物品 20 3-3-1 孿生神經網路 20 3-4 場景遮罩 21 3-4-1 OpenCV遮罩(Mask) 21 第四章 研究方法 23 4-1 問題描述 23 4-1-1 情境與問題描述 23 4-1-2 目標 23 4-2 問題與挑戰 23 4-3系統架構 24 第五章 實驗分析 36 5-1環境配置 36 5-2 實驗數據 37 5-3實驗結果 37 5-3整體系統評估 41 第六章 結論 42 參考文獻 43 圖目錄 圖 1、研究目標 3 圖 2、節點識別跌倒(出自研究[2]) 7 圖 3、人體2D骨骼圖轉3D骨骼圖(出自研究[4]) 8 圖 4、跌倒檢測方法的總體結構(出自研究[4]) 8 圖 5、姿勢估計的跌倒辨識(出自研究[7]) 9 圖 6、用於人體檢測和姿勢分類的框架(出自研究[8]) 10 圖 7、姿態檢測結果(出自研究[10]) 11 圖 8、整體架構(出自研究[10]) 11 圖 9、 MMEFD 框架(出自研究[11]) 12 圖 10、YOLOv4結合灰階圖判斷是否跌倒(出自研究[12]) 13 圖 11、YOLOv1模型架構圖(取自[1]) 16 圖 12、YOLOv5模型架構圖 17 圖 13、 YOLOv7模型架構圖 18 圖 14、RGB影像轉深度影像圖 19 圖 15、孿生神經網路架構圖 21 圖 16、遮罩圖轉換 22 圖 17、系統架構總覽 25 圖 18、居家影像深度圖轉換及遮罩流程圖 26 圖 19、居家影像RGB圖轉深度圖-訓練期 27 圖 20居家影像RGB圖轉深度圖-使用期 27 圖 21、影像遮罩處理-訓練期 28 圖 22、孿生神經網路比對流程圖 29 圖 23、YOLO物品標籤-訓練期 30 圖 24、YOLO物品標籤-使用期 30 圖 25、YOLO危險物標籤-訓練期 31 圖 26、YOLO危險物標籤-使用期 31 圖 27、孿生神經網路比對-訓練期 32 圖 28、孿生神經網路比對-使用期 33 圖 29、LineBot連結ChatGPT進行問答 34 圖 30、創建Linebot機器人-使用期 35 圖 31、RGB圖轉換深度影像之混淆矩陣 38 圖 32、深度影像遮罩之混淆矩陣 39 圖 33、IOU在不同數量影片之數值 40 圖 34、孿生網路的比對精確率 41 表目錄 表 1、相關研究比較表 14 表 2、訓練環境配置 36 |
參考文獻 |
參考文獻 [1] J. Redmon, S. Divvala, R. Girshick and A. Farhadi, "You Only Look Once: Unified, Real-Time Object Detection," 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR), Las Vegas, NV, USA, 2016, pp. 779-788 [2] ]P. -C. Chen, C. -H. Chang, Y. -W. Chan, Y. -T. Tasi and W. C. Chu, "An Approach to Real-Time Fall Detection based on OpenPose and LSTM," 2022 IEEE 46th Annual Computers, Software, and Applications Conference (COMPSAC), Los Alamitos, CA, USA, 2022, pp. 1573-1578 [3] Z. Cao, G. Hidalgo, T. Simon, S. -E. Wei and Y. Sheikh, "OpenPose: Realtime Multi-Person 2D Pose Estimation Using Part Affinity Fields," in IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 43, no. 1, pp. 172-186 [4] Steven J. Nowlan, Geoffrey E. Hinton. Simplifying Neural Networks by Soft Weight-Sharing[J]. Neural Computation,1992,4(4):473-493. [5] Yoshua Bengio, Patrice Simard, and Paolo Frasconi. “Learning long-term dependencies with gradient descent is difficult”. In: IEEE transactions on neural networks 5.2 (1994), pp. 157–166 [6] Ziwei Chen, Yiye Wang, Wankou Yang, “Video Based Fall Detection Using Human Poses”. In Arxiv:2017.14633 [7] Y. -H. Liu, P. C. K. Hung, F. Iqbal and B. C. M. Fung, "Automatic Fall Risk Detection Based on Imbalanced Data," in IEEE Access, vol. 9, pp. 163594-163611, 2021 [8] T. -H. Tran, D. T. Nguyen and T. Phuong Nguyen, "Human Posture Classification from Multiple Viewpoints and Application for Fall Detection," 2020 IEEE Eighth International Conference on Communications and Electronics (ICCE), Phu Quoc Island, Vietnam, 2021 [9] J. Macqueen, “Some methods for classification and analysis of multivariate observations,” in Proceedings of Berkeley Symposium on Mathematical Statistics & Probability, vol. 1, pp. 281-297, 1965. [10] P. Chutimawattanakul and P. Samanpiboon, "Fall Detection for The Elderly using YOLOv4 and LSTM," 2022 19th International Conference on Electrical Engineering/Electronics, Computer, Telecommunications and Information Technology (ECTI-CON), Prachuap Khiri Khan, Thailand, 2022 [11] D. Ros and R. Dai, "A Flexible Fall Detection Framework Based on Object Detection and Motion Analysis," 2023 International Conference on Artificial Intelligence in Information and Communication (ICAIIC), Bali, Indonesia, 2023 [12] R. Li, D. Li and M. Zhang, "A Real-time Fall Detection System Using ToF Depth Images," 2022 IEEE 10th International Conference on Smart City and Informatization (iSCI), Wuhan, China, 2022, pp. 41-48, doi: 10.1109/iSCI57775.2022.00016. [13] A. Bochkovskiy, C.-Y. Wang, and H.-Y. M. Liao, “Yolov4: Optimal speed and accuracy of object detection,” arXiv preprint arXiv:2004.10934, 2020. [14] Y. Zhou, W. Zhu, Y. He and Y. Li, "YOLOv8-based Spatial Target Part Recognition," 2023 IEEE 3rd International Conference on Information Technology, Big Data and Artificial Intelligence (ICIBA), Chongqing, China, 2023, pp. 1684-1687 [15] Y. Lecun, L. Bottou, Y. Bengio and P. Haffner, "Gradient-based learning applied to document recognition," in Proceedings of the IEEE, vol. 86, no. 11, pp. 2278-2324, Nov. 1998 [16] Chien-Yao Wang, Alexey Bochkovskiy, Hong-Yuan Mark Liao, “YOLOv7: Trainable bag-of-freebies sets new state-of-the-art for real-time object detectors”, arXiv:2207.02696 [17] J. Mueller and A. Thyagarajan. Siamese recurrent architectures for learning sentence similarity. In Proceedings of AAAI, pages 2786–2792, 2016. [18] I. Culjak, D. Abram, T. Pribanic, H. Dzapo and M. Cifrek, "A brief introduction to OpenCV," 2012 Proceedings of the 35th International Convention MIPRO, Opatija, Croatia, 2012, pp. 1725-1730. |
論文全文使用權限 |
如有問題,歡迎洽詢!
圖書館數位資訊組 (02)2621-5656 轉 2487 或 來信