Event-based supervision in different room layouts
A disadvantageous camera angle might prevent a scenario to trigger or the camera might have a poor person detection rate. This document describes how the room layout affects the event-based supervision performance.
Camera at the end of the bed is a good placement
Placing the camera at the end pointing along the bed will give good performance of the camera analytics, since the all sides of the bed are covered.
Camera successfully detects moving in and out of bed
Camera detects falling out of bed both with Out-of-bed and Person-in-room
Camera at an angle from the bed might be problematic
Placing a camera at an angle might result in the person’s bounding box being larger than the bed and the mask (the yellow outlines in the videos on this page), not being triggered. It is therefore important that the scenarios are tested to verify that there is no issues with the camera placement.
Detecting getting into bed fails since the bounding box is outside the bed.
The issue with camera angle is reduced if the distance is large enough.
The mask is triggered since the distance to the bed is far enough.
Falling out of bed is detected since the distance is far enough.
Placing the camera close to the bed should be avoided
A shorter distance from the camera will result in issues with the proportion of the object versus the bed size.
Detecting getting into the bed fails since the bounding box is outside the bed.
Avoid camera placement where the person can be partly covered
Object Analytics might not detect the person if he/she is partly covered, especially when using the IR mode at night. This since the video is gray toned, something that results in reduced contrast of the video stream.
Falling out of bed is not detected since the person’s bounding box does not leave the bed before the analytics looses track of the person. This is caused by the person being partly hidden behind the bed.