You can choose between different learning models, depending on the scene you want
analyzed, so that changes in the scene can be properly detected.
What you should know
- Each learning model is optimized for different types of scenes. The scene
background is analyzed and learned so that differences in the scene are
detected. When you change the learning model or the motion analysis mode, or
click Relearn, the background is relearned. The
relearning process can take up to a few minutes, depending on your configuration
and learning models.
- You can enable the Show scene background or the
Show motion analysis option to visualize the
sensitivity of your configuration. Detected differences from the background are
highlighted in blue if they are minor, in green if they are moderate, or in red
if they are severe.
NOTE: Depending on which scenario type you are configuring, some
settings might be unavailable or hidden if Advanced mode
is
disabled.
Procedure
-
From the Config
Tool home page,
open the Video task.
-
From the area view, select a camera that you applied an analytics scenario
to.
-
Click , and then click the scenario you want to configure.
-
Click the Foreground tab.
The background is learned using the default learning model.
-
In the Scene background section, select a learning
model:
- Color-based model
- This model is designed for indoor scenarios and thermal cameras. This model
continuously learns the colors of the scene and compares them to the current frame.
- Edge-based model
- This model is designed for outdoor scenarios. This model continuously learns the edges
of objects in the scene and compares them to the current frame.
-
- Startup learning period
- Specifies the duration for which the scene is analyzed
when the learning model is changed or when you click
Relearn.
NOTE: The initial scene should not contain any moving objects in the
foreground during the startup learning period because the object
detection model will learn to consider them as part of the
background.
- Continuous learning period
- Specifies the duration for which the scene is analyzed
continuously in the background.
- Reference image
- This model is designed for detecting changes in empty areas with unchanging lighting,
such as in an elevator. To ensure that the scene you want is learned, click
Relearn to manually capture the reference image when the scene
is empty. The current frame is compared to this reference image.
-
In the Motion analysis section, select the type of changes
that you want to detect:
- Detect all changes
- People and objects that are detected are considered part of the foreground, and can
trigger alerts, whether or not they are moving.
NOTE: When this option is selected, the
Show motion analysis option is not available.
- Detect stationary objects
- Only people or objects that remain still for the specified Stationary
object time period are considered part of the foreground, and can trigger
alerts.
-
Click Apply.