We configure two ROIs for this use case’s front-facing outside camera scene. The larger blue rectangle was called the Outside Front View, and the smaller blue rectangle closest to the vehicle was named the Proximity Area - Front. We then defined rules for each of these zones.

For the larger Outside Front View zone, we attached two rules; one for vehicle detection and one for person detection. The details of the rules are similar to the ones we have already defined above, so the UI is not included here.

For the Proximity Area - Front zone, we defined a rule called the Proximity Alert where we use the Zone Counting rule type but this time with detection of seven different object types including well-defined classes that include person and motorcycle, plus a generic Moving Object class to include anything that is not already included in the available object classes. We used a Reduce processing type to collect a simple count every second.

The camera mounted outside on the rear of the bus with a view of the right-hand side was also configured with two ROIs. The Outside Right Zone is configured with the same rule as the front-facing outside camera rule.

The Outside Right Zone is configured with the same rule as the front-facing outside camera rule with object detection of persons and vehicles.

For this view’s outside rear proximity detection rule, we used a Line Crossing rule type with the same object detection classes as the front proximity rule. We used a Split post-processing type to collect both the class name and dwell type for any objects crossing the virtual line toward the back of the bus.

Our configuration for the Outside Left Zone and Left Proximity Area Zone is identical to the right side. So we only show the ROIs defined in the following screenshot but omit the rule definition UI screenshots.