IP攝影機即時影像之跌倒偵測法——使用加速度特徵

Automatic Computed Tomography Image Quality Assurance Analysis

跌倒是老年人或病人時常發生的意外，而且也容易造成重大傷害。醫院中為預防跌倒發生常投入許多醫護人員的注意，也因此提升了營運成本。所以本研究將利用現有的監視裝置，結合跌倒偵測與簡訊通知的功能來輔助醫護人員。我們利用 MATLAB Simulink 和影像處理技術，設計一套具準確性的 IP 攝影機 （IP camera） 即時影像跌倒偵測系統。同時採用實驗的方式，蒐集多位受測者行走跌倒的即時影像，進而找出受測者處於跌倒動作時的加速度特徵來進行跌倒偵測。為了提高跌倒偵測的正確率，本研究結合長寬比及加速度來偵測跌倒。實驗證明在未結合加速度偵測法時之正確率為 86.67%，經結合加速度偵測法後其正確率提升為 93.33%。

The fall-accident is often occurred at the elderly or patients and the accident will cause a significant harm. To prevent the falls, hospitals always put a lot of medical staffs to pay attentions to their patients and thus increase the operating cost. Therefore, we employ the camera devices and combine falling-detection and message-notification functions to help the medical staffs. This paper proposes a real-time video falling-detection system that is implemented by using the MATLAB Simulink platform capture the video from the IP camera. Simultaneously, we take a lot of experiments to capture a number of walking-falling videos and then analyze the acceleration character for various falling actions. In order to improve the accuracy of fall detection, this paper combines the aspect-ratio and acceleration for the falling-detection. According to the experiments, the accuracy of falling-detection using aspect-ratio only is 86.67%, and the accuracy will be promoted to 93.33% by integrating the aspect-ratio and acceleration algorithms.