I can't figure out how you'd check the tips of the wings, but the leading edges should be visible from up by the nose. Similarly, the trailing edges from the tail.
It's also easy to be extra cautious. bias to 0 false positives. Sure it will detect problems that turn out to be dirt or poor lighting or something, but one visual check seems faster than 20 (or however many are necessary).
in a controlled environment like that, you could require stuff to be painted computer vision friendly colors. Even more, i'd bet oil reflects differently than metal under different lighting. internal stuff could be lit up with IR or some other part of the spectrum for easy detection.
I dunno. seems like perhaps a few pounds of electronics.
Sure. Maybe I shouldn't have picked the engine section as the example.
In the context of this thread, way more than that is needed.
> Put another way, the latest fighter jets and commercial heavies still have significant visual preflights.
And that is fine. The stakes are high, the plane itself is much more complex and placed in high demand, plus they can hire staff.
A plane lands, taxis to the gate. Passengers disembark, the plane gets cleaned and refueled, people on the ground checked it out, new passengers board (me), and it taxis from the gate. At no point did the pilots leave the airplane to "walk-around".
Other examples–checking the oil, the tank sumps, the intake, the tires, the flap tracks, the rudder bolt(s), the wing tips, the tailwheel springs, etc.–are the stuff that software largely can't replace, and are the more time-consuming part of taking off.
Put another way, the latest fighter jets and commercial heavies still have significant visual preflights.