擺動鏡頭攝影機於跌倒偵測之研究

A Study of Swing-Lens Cameras Applied in the Fall Detection

隨著高齡化社會的衝擊，老年人的照顧議題越來越受到重視，其中跌倒為最常見的問題，並會對身理及心理造成嚴重的傷害。因此，許多研究者皆提出各種不同的影像跌倒偵測系統，以達到及時發現跌倒之目的。但大多數此類系統因為皆採用固定鏡頭的監測方法而無法普遍被使用，固定鏡頭不但限制了系統的監測範圍，亦降低了系統的實用性。本研究使用Visual Studio C++配合OpenCV開發出一套擺動鏡頭跌倒偵測系統。這套系統透過計算影像重心位移時之加速度、斜率和影像長寬比來即時判斷跌倒行為的準確性。有意義的是，偵測系統的實驗結果和跌倒行為的模擬有高度的符合。本研究實作之系統對於跌倒偵測之平均正確率達86.66%，對於鏡頭擺動追蹤之正確率達90%。

Come with the impact of aging population society, elderly care issue is getting more attention. One of the common problems in elderly care is the tumble (fall), which would lead to a severe injury on physiology and psychology. Therefore, lots of researchers have proposed a variety of image detection systems for identifying the fall events in time. But those monitoring methods cannot be commonly used in elderly care due to detect camera with fixed-lens (no swing-lens). Fixed-lens did not only limit the scope in monitoring but also reduce the usability of the system. In this study, we develop a fall detection system with the swing-lens camera by using Visual Studio C++ and OpenCV. This system can accurately judge whether fall or not after calculate the acceleration, slope and aspect ratio of the mass center movement in the scene. Significantly, experiment results were highly comparable with our fall scenario simulation. Average accuracy rate of fall detection in the proposed system is 86.66% and the accuracy rate of lens tracking is 90%.