AI-SPILL


The performance is not adequate, what else can I configure?


Go back


Installation Constraints

To ensure a proper functioning of AI-SPILL the camera should be positioned so to have a perfect view of the whole area to be monitored. This requirement is quite strict as it is very likely that some elements might occlude the view and lead to failed detections: due to the fact that this application is mostly used in hospitals or private homes, for instance beds or tables could cause this situation.
People framed should, anyway, take up at least 20x20 pixels in height.
N.B. To ensure a proper functioning of AI-Spill you have to set the correct values for the camera calibration and calibrate the algorithm.

AI-Spill

WARNING: Please read the disclaimer to make sure that all of the installation constraints have been fulfilled!

Go back


Configuration

AI-Spill is the integrated video analysis solutions of A.I. Tech that allows to recognize dangerous situations such as patient falls. Early intervention, guaranteed by the immediate alarm generated by the module, could be crucial to the patient's life.

WARNING: AI-Spill expects the alarm to be generated only if there are fewer objects in the scene than those specified with the parameter “Maximum number of objects in the scene”. This is because the automatic system is designed to detect falls when the person is alone. If there are more objects than those indicated, the text SAFE MODE appears on the image and alarms are disabled.

Spill sensor configuration

Rilevamento Cadute

In the drop-down menu on the left, by selecting “Algorithm parameters” and then “Spill detection” you can add, remove and edit Spill sensors within the scene framed by the camera.
Specifically, by clicking on “Add sensor” you can draw a virtual coil inside the image through “Redraw the sensor”, until a coil of desired dimensions is drawn.
For each sensor it is possible to specify:

  • Sensor ID: identification of main sensor
  • Sensor Name: unique name of the associated sensor and used in events generation

Falling detection parameters configuration

Falling detection parameters
0.5
0.5
0.5
0.9

In the drop-down menu on the left, by selecting “Administrator parameters” and then “Falling detection parameters” you can define the parameters of the fall detection.
The parameters in this section are:

  • Moving confidence: a large value allows you to detect a moving object more easily, while a value too small may cause failure to detect.
  • Moving latency: minimum time, expressed in seconds, before a moving object can be detected.
  • Unmoving confidence: a large value allows you to detect a still object more easily, while a value too small may cause failure to detect it.
  • Unmoving latency: minimum time, expressed in seconds, before a still object can be detected.
  • Falling man confidence: a large value allows you to detect a fallen man more easily, while a value too small could cause failure to detect it.
  • Falling man latency: minimum time, expressed in seconds, before a fallen man can be detected.
  • Maximum number of objects in the scene: if the number of objects exceeds this threshold, the algorithm passes in SAFE STATE.
  • Minimum unmoving object overlap: the minimum percentage of similarities between successive instances to consider the object as still.
  • Unmoving distance: minimum distance between successive instances to consider the object to be stopped.
  • Sampling time: Minimum sampling time in ms between processed frames.
  • Form ratio falling man: maximum shape ratio (h / l) to consider it as a fallen man.
  • Deviation from ideal rectangle: minimum percentage of deviation from the ideal rectangle to consider it as a fallen man.

Go back