The VE350 is designed to detect left or removed items in moderately to less busy outdoor environments. Typical deployments include commercial buildings, where items are not to be left behind or removed, as well as less busy locations at airports, bus stations, and so on.
The VE350 is designed to work in outdoor situations. It will not work well in extremely busy environments. Camera position should normally be one or two storey's off the ground, and the camera angle can be flexible, depending on the field of view you wish to capture.
Select Server
> Configuration.
Select a camera from the Devices
branch of the left pane.
In the Analytics Engines
tab, select VE350. By default VE250 is selected, in which case you must
first deselect VE250 (see how
to select analytics for more help).
In the Analytics Configuration
tab, select Intel_VE350 from the
Analytics Engines combo box.
Select the Masks
tab.
Adjust the Length of
time (seconds) value to reflect how quickly the video engine
should alarm on left or removed items.
On the Persistence
Mask tab, adjust the yellow mask as required. Select
the Erase radio to erase the yellow
mask, and select Draw to draw
the mask. The Size slider adjusts
the pen thickness.
Click the Perspective
tab and accurately set the scene perspective.
Click the Alarm size
tab to define the minimum and maximum object sizes which are alarm candidates.
Optional: Select the Advanced
tab to configure additional settings. Usage:
|
Background Model |
Different background models do different things. We will continuously develop different models to meet the customer needs. Our current models include: Static: This is to be used in situations where the background is relatively static(no periodic movement in the background such as swaying trees). This is the fastest of our background models. Dynamic: This is to be used in situations where there are periodic moving objects (like a tree branch moving back and forth in a fairly constant wind, or caused by unstable camera mounting). Though this background model can handle both static and dynamic backgrounds, it uses more CPU power than the Static model and thus not recommended for static backgrounds. |
|
Update Background (seconds) |
Time to learn the background image. This entry determines the approximate length of time it will take a foreground object that becomes stationary to get merged into the background. NOTE: if an object has a periodic movement with a period greater than the specified time, then that object will always be considered a foreground object (never merged into the background). NOTE: The background is learned based on the last Update Background Seconds of video. The background is updated all the time, not only when the analytics engine starts. |
|
Sensitivity |
Sensitivity to sudden appearance of objects. High sensitivity (slider right) means more alarms appear because sudden appearances of objects are allowed. Low sensitivity (slider left) means less alarms appear because we need to ensure the item was dropped off by someone. The detected foreground objects are classified as alarms or non-alarms. Sensitivity controls how many detected objects are classified as alarms but doesn’t control the number of detected objects. Warning: the higher the sensitivity (slider right), the higher the chance of false alarms. |
|
Low Contrast |
A foreground object is assumed to have some contrast with the background (ie different colour/intensity from the background). By enabling low contrast, you assume that the contrast between foreground objects and the background objects are much smaller than the default settings. Since contrast is much smaller a different detection algorithm is used to ensure that objects with lower contrast to the background are detected. This checkbox controls the number of objects detected where as sensitivity slider controls the number of objects classified as alarms. Warning: by checking this checkbox the number of false alarms may increase at the same time low contrast objects will be detected and alarmed on. |
|
Use edge information |
Uses edges to reduce false positives. |
|
Reduce false alarms from people |
Attempt to filter people to further reduce false alarms |
|
Merge stationary rectangles |
If stationary objects overlaps, merge them into one rectangle. For example, someone drops a box, another box gets dropped which overlaps this box, both objects will be merged if this feature is checked. |
|
Enable MIM |
Leave on by default. |
Click OK to save settings.
Configuring the VE350 (see above) is required before creating a Rule. To configure a Rule using the VE350, follow the steps below:
Mark
the alarm zone using the Draw
and Erase radio buttons. The Size slider adjusts the pen thickness.
Alarm conditions may be set such that an alarm is raised only if 1) a certain number of pixels change in a certain time period; 2) a certain number of alarm-able events occur in a certain time period;
Two such conditions can be configured, denoted by "Alarm only, if" and "OR, at least".
For each Pixels Change condition that is enabled, a separate image mask will be added to the Current view shown below dropdown box. Use this dropdown box to switch between these views.
A Pixels Change condition mask will contain a green box representing the number of pixels specified. You can alter the numeric value, or you can size this green box by dragging its corners, which in turn will modify the numeric value.
Click Next to continue to next step in wizard.
Currently, there is no distinction between left and removed items. Leave both selected.
Ensure the correct frame rate is being utilized, and that enough CPU is available on the system to perform proper analysis. If problems still persist, contact Aimetis for support. Have sample video showing problems available for Aimetis support specialist to review.
See Also