AI-ATM

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


Installation constraints

A zenith camera pointing straight to the area is required to ensure proper functioning of AI-ATM Localization accuracy is higher when the effects of perspective are lower.
N.B.: To ensure a proper functioning of AI-ATM you have to to fill in the correct calibration values of the camera. Indeed, to configure AI-ATM you have to take all the steps required to configure AI-LOITERING sensors and AI-OVERCROWD sensors first and then, when possible, exclusion areas, in the section "Algorithm Parameters".

AI-ATM

Go back


Entrance area sensor configuration

Area Ingresso

In the drop-down menu on the left, by selecting "Algorithm Parameters" and the tab “Entrance areas" you can add, remove and edit entrance areas inside the scene framed by the camera. Specifically, by clicking on “Add sensor” you can draw an area of interest directly on the live image on the left.
The area of interest is represented by a polygon whose number of sides is virtually unlimited.
Once defined entrance areas, the objects detected inside the scene will be taken into consideration only when they cross one of these areas.


Loitering sensor configuration

Sensori Loitering

In the drop-down menu on the left, by selecting "Algorithm parameters" and the tab “Loitering sensor” you can add, remove and edit areas of loitering detection 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. After having added a sensor you can configure the following parameters:

  1. Sensor ID: unique identification of added sensor
  2. Sensor name: name associated to the added sensor

Besides these two parameters you can also configure the following to achieve an even better configuration:

  1. Confidence: especially in crowded scenes, the systems allows the setting of the confidence of an alarm. Basically the application generates an events if detects the presence of an object or a person for a certain amount (confidence) of the persistence time.
  2. Inhibition: to avoid the constant generation of alarms, the system allows to determine an inhibition timespan, which is the minimum interval of time between two consecutive events generated by the same object.
  3. Latency (s): it is the time of persistency of the object or the person inside the area of interest before the alarm is generated.


Sensor for crowd estimation configuration

Stima Persone

In the drop-down menu on the left, by selecting "Algorithm parameters" and then “Crowd estimation” you can add, remove and edit sensors for crowd estimation 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:

  1. Sensor ID: identification of main sensor (only allows whole numbers and it is not necessary to edit the default setting)
  2. Sensor Name: unique name of the associated sensor and used in events generation
  3. Event Period: interval of seconds between two consecutive events to be sent to an external server
  4. Person Area: draw a rectangle around a person of known dimensions located at the center of the scene so to properly re-determine the dimensions of the same person in each spot of the scene.

N.B.: To ensure proper functioning of AI-Crowd you should fill in the calibration parameters of the camera first.


Overcrowd sensor configuration

Affollamento

In the drop-down list on the left, by selecting "Algorithm parameters" and then “Overcrowd” you can add, remove and edit sensors for crowd estimation within the scene framed by the camera.

To get AI-Overcrowd results select the field “Enable Overcrowd”. Here, you can also set the following parameters:

  1. Confidence: A small value (< 0,5) makes the algorithm extremely sensitive, while a value that is too high (> 0,8) might lead to missed detections. It is advisable to use a value between 0,5 and 0,75.
  2. Inhibition: Inhibition time in seconds after an alarm is generated by the specific sensor.
  3. Latency: Minimum time of overcrowding (number of people is higher than a certain threshold) before the alarm is generated.
  4. Overcrowd threshold: If the number of people in the region of interest is above a certain threshold, the application generates an overcrowd alarm.

N.B.: To ensure proper functioning of AI-Overcrowd you should fill in the calibration parameters of the camera first.

Go back