I agree with @Antonius that I think that car driving past is involved.
Part of the problem with the Outdoor cam is that the actual INITIAL trigger is technically not caught on video. The initial trigger WAKES up the WCO, and then it captures what happens immediately proceeding what triggered it. It could have been a bird at the edge of the screen or any number of other things that triggered the PIR to screen the following few frames to see what may be of interest.
Additionally, it is unknown whether the PIR range may slightly extend beyond the range of the video screen (ie: slightly farther to the left or right).
The detection zone SHOULD act as a second constraint though. If you play the video in 1X speed there appears to be NO movement at all anywhere except for the car in the corner. HOWEVER, if you play this on a computer monitor (Mine is 17") and make the video full screen, then instead of playing it in 1X speed, you PULL the playback cursor back and forth fairly quickly, you will suddenly see there is plenty of movement and pixel changes. Mostly the shadows move within the detection zone when you move the playback slider really quickly, but also outside of the detection zone I saw the flag wave and the plants next to the flag wave, and the plants in the top left I saw a slight position change in the pixels, but most of those were outside the detection zone and wouldn’t have counted…only the slightly moving shadows on the driveway…I only mention the other movement to indicate there WAS some very slight movement, even though it is hard to see unless you blow it up and move the slider quickly.
I agree that all that this small movement and lighting changes SHOULDN’T normally matter. But, on the highest sensitivity it is possible that it might be enough to count as sufficient pixel changes to cause the camera to upload a video and check. Once the video uploads, then the AI screens the ENTIRE video, not just the detection zone areas (though Shawn said they are looking at an update so the AI only scans the detection zone rather than the full video), and so the car would then be detected.
So here is my primary theory: The cars are setting off the PIR sensor (because the PIR checks EVERYWHERE it can, it can’t be limited by a detection zone), and sometimes my PIR on my WCO picks up vehicles fairly far away because of their heat signature difference. Then the camera sees something in the detection zone (slight shadows that are so slight we can’t tell as easily with our human eyes…but you might see a little better if you move playback really fast). So a video is created and uploaded and the AI analyzes the entire thing and identifies a vehicle (whether parked in the driveway or driving on the road) and alerts to the detection of a vehicle.
Also, could be that the detection zone is a little close to the road. Maybe it even picked up the car. I know my detection zones seem to have a little bit of leeway for some reason. Maybe tilt the camera slightly more to the left, or block out the ones I marked in 1 and test if that makes a difference, and then try to block out 2 as well and see if that helps.
Also, possibly lower the sensitivity range a little. Maybe from 100 down to ninety something and move up and down until you see if that is helping at all while still picking up anything that enters your property.
Also, possibly lower the PIR range so that fewer cars on the road trigger it.
Basically a few experiments to find the correct sensitivity and zones for your particular situation. I had to do a few trials like this for my WCO until I found the what worked best…each situation is a little different.