| dc.description.abstract | In these decades, the economic situation and the living quality of human are quickly improved. Human life is consequentially lengthened. The long life brings a side effect; it is that the old-man disease is heavily increased. Thus people pay more attention to these kinds of health care. However, some kinds of accidents are occurred due to the negligence of nurses, the patient’’s conditions are more deteriorated. To reduce the deterioration, we propose an environmental monitoring system for such kinds of health care in this thesis. There are four kinds of monitoring in the proposed system : (i) patient fall down and crouch, (ii) leaving bed without permission, (iii) going out ward without permission, and (iv) walking trajectory.
In this study, we use an omni-directional camera to capture images for monitoring. The camera can easily take 360-degree surround images. In the fall down and crouch detection, we use principal component analysis to detect the main direction and length of the personal body, and then determine the patient situation. Another method is using the change of the height of the patient trapezoidal bounding box. In the leaving bed detection, we detect whether the patient leaves the pre-defined region or not. However, the criterion tends to result in a unstable judgment; thus we need to check whether the patient really lies flat on the bed. In the leaving ward detection, we first define a linear equation to represent the bottom of a door; if the foot point of a patient is passing through the boundary, the patient is judged leaving ward. In the walking trajectory detection, the foot point of a patient is recorded frame after frame to construct the walking trajectory on the calibrated omni-directional images.
The detection rate of the fall down and crouch detection using principal component analysis is 93%, and the detection rate of that using the height of the patient trapezoidal bounding box is 92%; the detection rate of leaving bed is 95%; and the detection rate of leaving ward is 95%.
| en_US |