This sounds like a job for machine learning. Are there any systematic efforts underway to use Google Earth to search for out-of-place objects?
Google is watching your backyard from above seems like a PR disaster to its already troubled image.
I think my parents had an aerial photo of their land a few years before even TerraServer was came around, from some guy who'd flown over a big chunk of territory taking photos and then sold them to the land owners. I assume this was far from the only person doing that, all over the US. Pretty cool back then, when a clear shot from the air of the area you lived in wasn't something you could grab at a computer any time you wanted, or with a cheapish drone and a few spare minutes. And before everyone else could creep on your space in the same manner on a whim, for that matter.
There is a significant discrepancy between the imagery viewable in 3D mode and the stuff that's only 2D, because Google simultaneously captures high resolution aerial photography and Lidar, particularly around cities. You can see vast differences in urban or rural landscape sometimes because the stuff you see in Google maps can be taken as far as years apart from the 3D aerial stuff.
For example, you can see where the data source and rendering method change just outside the city of Winnipeg on the South end. Probably testable in many other cities simply by seitching between google maps and google earth.
Check out how poorly the top five did in just finding large vehicles in a kaggle completion explicitly about annotating satellite images: https://www.kaggle.com/c/dstl-satellite-imagery-feature-dete...
I get where you're coming from, but there will be (probably literally) a million abandoned cars for every car worth investigating.