After selecting cameras and models for the DG, you need to set up AI Tasks to monitor objects from these cameras.
To set up an AI Task, in Step 3: Setup AI Tasks to Monitor, click the Add (1) button.
The AI Task Configuration dialog will be opened. The first thing you need to do is:
- Naming AI Task (1)
- Select the camera (2) you want to monitor for this task
Next, you need to set up rules. A rule is a set of conditions that defines and triggers an event.
Rule Type
In order to create a rule:
- Click the (+) button
- Select Rule Type
Rule | Description |
Presence | Detect the object(s) present in the specified region. |
Path | Detect Object(s) travelling over a specific path defined by a Starting ROI and End ROI. |
Boundary Crossing | Detect the objects crossing a defined boundary and provide incremental counting for each object class. |
Static Count | Count the number of specified objects in the image (or ROI). |
Inclusive | Check if another object(s) is within the boundary of the first-tier objects found. |
Entry/Exit | Detect if selected objects are crossing Entry and Exit boundaries defined by the user. Make sure the crossing direction is correctly defined. |
Virtual Zone/Line (ROI)
After selecting Rule Type, you can add a Virtual Zone/Line (ROI) to monitor objects in specific areas/directions of the image or skip this step to monitor the whole image.
- From Virtual Zone/Line, select shape type (1)
- Click add (2) to add shape to the image
- Drag and drop the shape to adjust the region you want to monitor
- Right-click (3) on the shape to add/remove additional point
Object Acceptance/ Zone direction
Object Acceptance
Each object is encompassed by a bounding box. With the object acceptance feature, you can select the portion of the bounding box to compare with the Zone Direction for it to qualify as a detection.
- All Corners: all corners of the bounding box
- Centre point: The centre point of the bounding box
- Any corner: Any corner of the bounding box is
- Any Overlapping: Any overlapping between the bounding box and ROI
Zone Direction
Next, select a zone direction to determine the position between the Object Acceptance and the ROI that qualifies as a detection.
- Inside ROI: Object Acceptance is inside the ROI
- Outside ROI: Object Acceptance is outside the ROI
- Right side: Object Acceptance is coming from left to right
- Below: Object Acceptance coming from above to below
- Left side: Object Acceptance coming from right to left
- Above: Object Acceptance coming from below to above
- Both directions: Object Acceptance coming from both directions
Object List
Not all objects from the model need to be included in the event; you can filter the objects you want to monitor from the object list.
- Click Select All to select all objects in the models, OR
- Click each drop-down list to select an object you want to include
- Click Remove All to remove all selected objects
Pipeline option (optional)
The pipeline option will try to detect license plate (ALPR) or text (OCR) inside the object's bounding boxes, if any.
When you enable Configure Filter, the pipeline will become an additional condition for event generation, which means the object needs to contain/not contain specific texts or license plates to trigger an event
From the pipeline filter configuration, you can:
- Specify the list of text conditions (1)
- Use logic AND/OR between texts (2)
- Choose if you should log the detected objects and text that don't match the condition (3)
Rule Parameters
Depending on the selected rule, you can configure additional parameters at the bottom of the UI
Every Detection
Select the Every Detection type to generate events constantly when an object is in the ROI.
Object Appearing
Select the Object Appearing type to generate events only the first moment the object appears.
Object Disappearing
Select the Object Disappearing type to generate events only the last moment the object appears.
Duration
Select Duration type to generate events when objects stay in the ROI for a defined time.
Path
Generate events if the total amount of time an object moves from start to end ROI is in the duration limit. This rule will also incrementally count the detected objects over time.
Boundary Crossing
Generate events when a specific number of objects cross the ROI. This rule will also incrementally count the detected objects over time.
Static Count
Density mode: enable to calculate based on % ROI that the object covers
Limit mode: trigger if the number of detected objects meets the comparison limit.
Minimum duration time: the frequency (second) at which an event is generated
Inclusive
With the inclusive rule, the event is triggered depending on the space relativity between the main and inclusive objects.
First, you need to choose the main object from the object list (1) and add inclusive objects (2)
- Select proportion mode if you want the space relativity between the main object and the inclusive object to be proportional to each other.
- Select absolute mode if you want the space relativity between the main object and the inclusive object to be the exact same as you draw.
Then you can adjust the boundary (3) between each object on the image. This boundary is space-relativity between the two objects, not ROI.
Next, select the trigger option to control how events are generated
Every detection | The event is generated every time the inclusive object is within the boundary of the main object. |
When at least X inclusive objects are found | The event is generated when a specific number of inclusive objects is found within the boundary of the main object. |
Every time inclusive data changes | The event is generated once every time the number of inclusive objects changes. |
Save Rule
Once you are done, click Save Rule (1) to save your rule.
Rule Join Condition
Once a rule is saved, you can continue adding other rules for this event and join them with Rule Join Condition (1)