Tracker-Based Use Cases
VisionAI Observ provides more advanced vision pipelines for tracking people and vehicles in motion, allowing for temporal triggers versus the usual single-frame logic.
Event Descriptions
Multiple events fall under this category, including the following:
Person Dwelling - Utilizes trackers to check if a person has been idle in a frame for a prolonged period of time. Ideal for loitering detection.
Person Check-In/Out - Tracks the timestamps when a person enters and exits a specific zone. Useful for tracking customer service time and monitoring entry and exit of people in private property.
Pedestrian Flow Threshold Exceeded - Tracks when a specific number of people exit a zone in a predefined period of time. Useful for watching for stampedes and other cases where a significant number of people move from one zone in the camera to another.
Vehicle/Motorcycle Flow Threshold Exceeded - Definition.
Illegal Parking Detected - This use case is deemed confirmed by using a tracker to see if a vehicle is in an illegal parking spot for a prolonged period of time. Ideal for watching roadsides and private property for illegal parking.
Vehicle Check-In/Out - Tracks every vehicle that enters the area. Useful for monitoring entry and exit in private property.
Wrong-Way Driving - Utilizes a tracker across two zones to detect when a vehicle enters and exits in a direction opposite to the designated flow of traffic. Ideal for monitoring one-way streets, highway on/off-ramps, and restricted-access lanes.
Illegal U-Turn Detection - Utilizes a tracker across two zones to detect when a vehicle enters a zone, reverses direction, and exits back through the same or an adjacent zone against the permitted flow. Ideal for enforcing no-U-turn restrictions at intersections and medians.
Event Logic
VisionAI Observ tracks when an object enters and exits a defined area. Depending on the use case, either one or two zones can be configured in the camera view.
A single zone is sufficient when you only need to know how long an object stayed within one area. This covers cases like measuring wait time in a queue zone (e.g., checking how long customers wait in line), measuring dwell time in a store or on private property, or logging every vehicle that enters a private property for retention in a database.
Two zones are used when you need to track an object moving from one zone into another — entering and exiting Zone 1, then entering and exiting Zone 2. This sequence is what allows Observ to detect directional or order-dependent behavior, such as a vehicle counterflowing against traffic, a vehicle swerving where it shouldn't, or a person taking an illegal entrance.
Note that these use cases are templates based on logic, not strict definitions tied to their names. If a scenario doesn't exactly match the name of an existing use case but follows the same underlying zone logic (one zone for dwell/duration tracking, two zones for sequential entry-exit behavior), the corresponding template can still be applied to configure it.
Event Setup
Create your events as usual, optionally following this camera task creation tutorial. The one thing to note is to ensure GPU compute is enabled instead of CPU decode — this is set under the Advanced section during Task setup.
This can be configured either during initial task setup or modified afterward, for cases where a non-tracker-based task is created first and a tracker-based event is added later. To modify it after the fact, click into the task, open Settings, and adjust the Decoder setting under Task Information.
Once the task has been updated to use GPU decode, you can refresh the page and add new detection events like those defined above.
Last updated