跳到主要內容

臺灣博碩士論文加值系統

(44.211.31.134) 您好!臺灣時間:2024/07/21 06:15
字體大小: 字級放大   字級縮小   預設字形  
回查詢結果 :::

詳目顯示

我願授權國圖
: 
twitterline
研究生:王敬仁
研究生(外文):WANG,CHING-JEN
論文名稱:基於多模型裁決辨識架構之藥物辨識系統
論文名稱(外文):A Multi-Recognition Framework for Medication Identification Systems
指導教授:張榮貴張榮貴引用關係
指導教授(外文):CHANG, RONG-GUEY
口試委員:陳敬陳鵬升簡廷軒張榮貴
口試委員(外文):CHEN, JINGCHEN, PENG-SHENGCHIEN, TING-HSUANCHANG, RONG-GUEY
口試日期:2024-06-21
學位類別:碩士
校院名稱:國立中正大學
系所名稱:資訊工程研究所
學門:工程學門
學類:電資工程學類
論文種類:學術論文
論文出版年:2024
畢業學年度:112
語文別:中文
論文頁數:74
中文關鍵詞:藥物辨識人工智慧醫療安全
外文關鍵詞:YOLOv8NVIDIA Jetson Nanomedication recognitionartificial intelligencemedical safety
相關次數:
  • 被引用被引用:0
  • 點閱點閱:17
  • 評分評分:
  • 下載下載:0
  • 收藏至我的研究室書目清單書目收藏:0
醫療環境中的藥物給藥錯誤對病人安全構成嚴重威脅。本研究旨在應用人工智慧技術,特別是YOLOv8模型,來輔助護理人員正確進行藥物管理,從而降低藥物錯誤率。本論文與嘉義基督教醫院合作,開發了一套基於NVIDIA Jetson Nano開發板的智慧藥物辨識系統,此系統配備高解析度攝影機,能夠實時辨識達319種不同藥物。通過此系統與階層結構架構辨識方法,不僅提升了藥物辨識的準確性,也提高了護理作業的效率。
研究中發現,藥物影像在實際環境中常遭遇多種挑戰,例如光線不足、陰影遮擋及像素問題等,這些因素顯著影響了影像識別的效果。為了克服這些挑戰,我們探討並實施了多種圖像前處理技術,如自動曝光調節、陰影消除算法及像素增強技術,以優化藥物的影像質量,進一步提升模型的辨識能力。此外,進行了模型的參數微調,包括學習率的調整和訓練迴圈的優化,以適應藥物辨識的具體要求。
本研究不僅成功將YOLOv8目標檢測模型應用於藥物識別任務,也在真實的醫療環境中展示了AI技術的實用性和有效性。透過進一步的系統驗證與臨床試驗,本系統有望成為提升藥物管理安全性與效率的重要工具。此項研究為未來醫療領域中類似技術的發展和應用提供了堅實的理論和實驗基礎。
In a medical environment, medication errors pose a serious threat to patient safety. This study aims to apply artificial intelligence technology, particularly the YOLOv8 model, to assist nursing staff in correct medication management, thereby reducing medication error rates. In collaboration with Chiayi Christian Hospital, a smart medication recognition system based on the NVIDIA Jetson Nano development board was developed. This system is equipped with a high-resolution camera capable of real-time recognition of 319 different medications. Through this system and a hierarchical structure-based recognition method, not only has the accuracy of medication recognition been improved, but also the efficiency of nursing operations has been enhanced.
The study found that medication images often face various challenges in real-world environments, such as insufficient lighting, shadow obstruction, and pixelation issues, significantly affecting image recognition effectiveness. To overcome these challenges, various image preprocessing techniques were explored and implemented, such as automatic exposure adjustment, shadow elimination algorithms, and pixel enhancement techniques, to optimize the image quality of medications and further improve the model's recognition capabilities. In addition, parameter fine-tuning of the model was conducted, including adjustments to the learning rate and optimization of training loops, to meet the specific requirements of medication recognition.
This study not only successfully applied the YOLOv8 object detection model to medication identification tasks but also demonstrated the practicality and effectiveness of AI technology in real medical environments. Through further system validation and clinical trials, this system is expected to become an important tool for improving medication management safety and efficiency. This research provides a solid theoretical and experimental foundation for the development and application of similar technologies in the future medical field.
一、 緒論
1.1 研究背景
1.2 研究動機
1.3 研究目的
各章節介紹
二、 相關研究
2.1 Yolov8模型介紹
2.2 藥物辨識
三、 研究方法
3.1 藥物辨識面臨的挑戰
3.2 裝置與模型配置
四、 實驗結果
4.1 Mean Average Precision
4.2 訓練成本
4.3 開發板實際辨識速度
4.4 預測框損失
五、 結論
參考文獻
[1] https://www.grandviewresearch.com/industry-analysis/microprocessor-market
[2] You Only Look Once: Unified, Real-Time Object Detection,2015 Joseph Redmon, Santosh Divvala, Ross Girshick, Ali Farhadi
[3] CSPNet: A New Backbone that can Enhance Learning Capability of CNN,2019 Chien-Yao Wang, Hong-Yuan Mark Liao, I-Hau Yeh, Yueh-Hua Wu, Ping-Yang Chen, Jun-Wei Hsieh
[4] UAV-YOLOv8: ASmall-Object-Detection Model Based on Improved YOLOv8for UAVAerial Photography Scenarios,2023 Gang Wang, Yanfei Chen *, Pei An, Hanyu Hong, Jinghu Hu and Tiange Huang
[5] Liu, S.; Qi, L.; Qin, H.; Shi, J.; Jia, J. Path Aggregation Network for Instance Segmentation. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, Salt Lake City, UT, USA, 18–23 June 2018; pp. 8759–8768.
[6] Li, X.; Wang, W.; Wu, L.; Chen, S.; Hu, X.; Li, J.; Tang, J.; Yang, J. Generalized Focal Loss: Learning Qualified and Distributed Bounding Boxes for Dense Object Detection. arXiv 2020, arXiv:2006.04388
[7] Automatic Pill Identification from Pillbox Images - DE Madsen et al., 2013
[8] Analysis of Hu's Moment Invariants on Image Scaling and Rotation Zhihu Huang, Jinsong Leng Edith Cowan University 2010
[9] Epshtein, B., Ofek, E., Wexler, Y.: ‘Detecting text in natural scenes with stroke
width transform’. Int. Conf. Computer Vision and Pattern Recognition, IEEE,2010
[10] Accurate system for automatic pill recognition using imprint information, 2015 Jiye Yu1, Zhiyuan Chen1, Sei-ichiro Kamata2, Jie Yang1
[11] CNN-Based Pill Image Recognition for Retrieval Systems,2023 Khalil Al-Hussaeni , Ioannis Karamitsos , Ezekiel Adewumi and Rema M. Amawi
[12] A Computational Approach to Edge Detection ,1986 JOHNCANNY, MEMBER, IEEE
[13]https://medium.com/@bob800530/opencv%E5%AF%A6%E4%BD%9C%E9%82%8A%E7%B7%A3%E5%81%B5%E6%B8%ACcanny%E6%BC%94%E7%AE%97%E6%B3%95-d6e0b92c0aa3
74
[14] AFPN: Asymptotic Feature Pyramid Network for Object Detection,2023 Guoyu Yang, Jie Lei, Zhikuan Zhu, Siyu Cheng, Zunlei Feng, Ronghua Liang
[15] https://fixthephoto.com/focal-length-comparison.html
[16]An acurate Deep Learning–Based System for Automatic Pill Identification: Model Development and Validation,2023 Junyeong Heo ; Youjin Kang ; SangKeun Lee ; Dong-Hwa Jeong ; Kang-Min Kim
電子全文 電子全文(網際網路公開日期:20290624)
QRCODE
 
 
 
 
 
                                                                                                                                                                                                                                                                                                                                                                                                               
第一頁 上一頁 下一頁 最後一頁 top