Hacker News new | comments | ask | show | jobs | submit login
Airplane is tripping on Google Maps (maps.google.com)
114 points by mcs on June 13, 2010 | hide | past | web | favorite | 25 comments

interesting - i guess the rgb (and luminosity?) channels are captured in succession? was this captured by a satellite?

I think that is what happened here.

Google Maps uses, among others, images from the Ikonos satellite. Here are the specs: http://www.satimagingcorp.com/satellite-sensors/ikonos.html

It has a panchromatic sensor (greyscale) and red, green and blue sensors. Ikonos has a ground speed of 6.8 km per second, so even if the sensors are only a few centimeters apart, the images would need to be recombined at an offset. The offset would be tuned for static ground objects.

So I guess the ghosting is due to both the altitude and the speed of the airplane relative to the ground.

Because the direction of the offsets in the plane is at the plane's direction it is much more likely caused by the plane's speed than its altitude or the speed of Ikonos.

Or this plane happens to be travelling in the direction of the satellite?

EDIT: here's another plane with the same (grey, red, green, blue) sequence: http://maps.google.com/maps?oe=utf-8&ie=UTF8&gl=us&#...

The sideways offset seems to be roughly half as much as the offset in the direction of movement of the plane, so at least that component would be due to the sensor platform moving...

That plane is headed straight for PHX, and isn't too far away, meaning it's probably traveling a lot slower than the other.

Which agrees with what Luc said. Essentially he's saying that Ikonos has to factor out its own motion. Hence the only motion left that's not factored out is the plane's.

Not quite. Even if the plane had zero ground speed, and the Iknonos were offsetting the four images to account for its own motion, the effect would still appear. That's because the offsets that align features on the ground will not be the same offsets that align features on a nearer plane (no pun intended), due to parallax.

If it was a parallax issue, there would be fringes of colour on anything tall. (The plane is very low in that image, probably going to or coming from the runway just to the southwest)

That's definitely not Ikonos imagery, the imagery is aerial.

It's also not the sensor distance that contributes to the halo effect for some satellites, it's the capture time. Often, each band is captured one after the other. Some satellites, like Rapideye's, have several seconds between certain bands. For instance their B/R bands are (I believe) first/last bands captured (of 5 total), so all moving clouds in their imagery get a red/blue halo.

the resolution is to high for Ikonos. But apart from that you are right.

The specs site mentions resolution of 0.82 meters panchromatic - would that be per pixel? And if so, how come the images at Google Maps have higher resolution?

Google uses a variety of data sets, including regular aerial photography and multiple remote sensing products. If houses and cars look sharp at the highest zoom, you're probably looking at aerial photography.

Exactly, and the plane's image contains small objects above that resolution threshold, so it seems that is wasn't taken by the Ikonos system.

In the UK at least, the higher resolution aerial imagery is captured by aircraft, from a couple of companies. The one I know is Geoinformation Group, which has a good historic archive as well as a continuous survey of the country.

That's literally half a mile from where I grew up. If you zoom out a few times you'll see Lake Wylie. My mom still lives right there on Lake Wylie Drive right off of Mount Gallant.

What are the chances of that?

Given the US are approximately 10 M square km, the radius you define is .5 mile. Let's say that 1% of the US is populated like the area where looking at.

PI * 0.5 * (0.5 (square miles)) = 2.03417191 square km. let say 2. That's the size of the disc around the plane where your house could be.

10000000 * 0.01 * 2 = 200000 That's the total number of those discs fitting in densely populated US.

So 1 chance out of 200000, that's in the same ballpark of the size of HN community, so it was likely to happen to someone, but it had to be YOU ;)

Unrelated, because there are only O(N) possible location-hacker pairs, not O(N^2).

I have the suspicion that you have no idea what you are talking about and people are upvoting you for using landau-notation.

Actually, I used landau notation because it's the easiest way to explain why you're wrong.

I didn't use exact numbers because I have no idea what numbers you have to use to convince yourself that the number of hacker-location pairs would scale quadratically with the number of hackers, as in the birthday paradox. It makes absolutely no sense.

No kidding? I grew up in Clover, SC and thought the same thing. Such a small world.

Holy cow, if you go into street view on Edna Dr. under the plane you can see contrails, probably not the same plane but still kinda neat.

That's just plane gay...

Applications are open for YC Summer 2019

Guidelines | FAQ | Support | API | Security | Lists | Bookmarklet | Legal | Apply to YC | Contact