Results ELI5
Methodology
I modeled the drive from residential homes to 22 sensitive sites, including military bases, federal and intelligence buildings, a national laboratory, and the headquarters of defense contractors, to find out what share of people would pass a Flock camera and be observed on their way to work. I say observed rather than detected because a Flock camera does not need a license plate to find a car. It also records the vehicle's make, color, and other details, which is why the 75, 150, and 250 foot tests below ask how far from the road a camera can still pick a car out.
Across the 22 sites, the typical result is that two out of every three trips pass a confirmed camera. Eglin Air Force Base is the highest at more than nine out of ten. A confirmed round trip at the typical site passes 3.17 cameras on average. Over a working year of 250 trips, that is roughly 792 times the same car is recorded.
Explore the distance test
Choose a site, then switch between the Primary H0 distance test and the more cautious Robust H0/H1/H2 result.
Why four colors, not just "observed"?
A home can usually be reached by more than one road, and cameras sit at angles, so a plain observed/unobserved split would over- or under-claim. Each modeled route is checked for a camera the model can confirm by direction and corridor.
Passes a camera (Confirmed): the route passes a confirmed camera. In Robust mode this holds on every plausible road; in Primary H0 mode, on the one primary road. Avoids all cameras (Unobserved): the route passes no such camera. Might pass a camera (Possible): a camera is on the route but can't be confirmed either way — its direction or edge can't be resolved, or (in Robust mode) the plausible roads disagree. Couldn't determine (Unresolved): no complete route could be established.
H0, H1, and H2
H0: the first road-access guess. H1: another plausible road. H2: a third option where access is ambiguous.
75, 150, and 250 feet
75 ft: narrowest test. 150 ft: the default field of view. 250 ft: widest test. Range, monitored direction within 35°, and same-road evidence are required. Crossing, frontage, grade-separated, and oncoming roads do not automatically count.