- 學位論文
以電腦視覺技術開發病房事件監控系統
Surveillance of Clinical Ward Activities Using Computer Vision
摘要
隨著電腦的普及造就了電腦視覺技術的進步，利用電腦視覺技術可以為生活增加更多的應用。本論文使用電腦視覺技術即時偵測病人上下床以及跌倒行為。 因為一般家庭或醫院，親屬或護理人員不容易能夠二十四小時全天照顧，此系統將可提供一個安全照顧的環境，可偵測病人是否上下床，當有發生跌倒事件時，系統會發出警報，提示護理人員或是透過手機發送訊息告知遠方的親屬。 本論文以網路攝影機為基礎，結合一套簡單快速的方法，利用背景相減來偵測出感興趣之前景物，並利用一些影像處理的技巧配合人類行為分析，判斷病人是否上下床以及是否發生跌倒之行為。攝影機架設在醫院容易發生跌倒行為的區域，當發生跌倒行為時，能夠即時發現發出警示，並通知護理站或是通知遠端的親屬，藉此增加救援的黃金時間，使傷害降至最小。 本系統的設計相較於其他系統簡單，有處理速度快、偵測準確率高、可同時偵測多人、用於即時偵測更具效率等優點。實驗部份，則模擬了各種可能發生的動作，如站立、蹲下、坐下、躺下以及各個角度可能發生跌倒的姿勢，皆有不錯的成果，相信本研究對病房照護及監測會有很大的助益。
並列摘要
The increasing popularity of computers has advanced the progress and applicability of computer vision technology. This thesis adopts computer vision technology to detect people falling out of bed. Since looking after a person 24 hours a day is difficult, whether in a hospital or at home, the proposed system can offer a safe environment. An incident of falling out of bed causes the system to send out an alarm to the person responsible for looking after the patient, whether they are nursing staff or a relative, by cell phone. This thesis uses a webcam video to do background subtraction and image processing and human behavior analysis to identify incidents of patients falling out of bed. The webcam is erected in an appropriate place in a hospital, so that the system can notify users in real time when falling occurs, thus increasing the opportunity for timely rescue, and reducing the possibility of injury. The proposed system is simpler than previous systems. It has fast processing, high detection and examination rates, and high efficiency. An experiment it performed by enacting various motions, including standing, squatting, sitting down, lying down and postures leading to falling. Experimental results demonstrate that the proposed system performs well, and effectively monitors patients.