I get it but good luck. Is that road heavily traveled at night? Can you not design a couple of MD blocks that wont "see" the road?
====
Example:
One would think IVS would be triggered for both of these targets. Nope.
(Z4 is running 0-5 Exposure, IR on Auto at night - which in this case is identical to manual 100%).
The Overexposure control is aggressive but I've come to like it vs white out of faces.
There is a shit ton of IR on that scene from 3 cameras. The distance is 45ft from the Z4, there's a 5442 bullet directly under it focused on the driveway but pointing same direction, and a mini-ptz parked looking at that same spot for LPR at night. The Z4 is dialed back on Exposure to prevent washout of faces.
Without
MD at night only I would miss most nighttime events
(Yellow on timeline is MD, Red is IVS/AI)
I spent
a lot of time for many months trying to get IVS to activate better at night on that scene and the best I could get was maybe 50% activation, but with faces washed out or blurred. Simple MD for nighttime hours solved the problem.
*So Yes I'm running Metadata currently, but I ran plain Tripwire for a very long time, results identical as far as target acquisition.
**I have two other cameras (4K-X and 5442 bullet) that see the street and DO activate IVS at night so I could and did for a while just blow off nighttime alerting on that zoomed in camera as I could easily determine a time to manually look for based on the other two cameras activation.
So why do the 4K-X and 5442 bullet activate IVS on this area at night and the zoomed in Z4 wont? I'm pretty sure it has to do with time it takes for AI to recognize and process a target AS WELL AS the composition of the scene. When zoomed in this tight, the Z4 has less time to react on a target entering the AI zone, and the overall scene is such that it has a hard time differentiating between static and moving objects in its FOV as well as identifying them as human/vehicle. Extreme contrast can help, as the expense of the image quality.