▲ 85 points
back
34 comments
Random thought: If you stick a very small adhesive mirror to the outside window before boarding, then footage "looking down" at the regularly-spaced railroad ties could contribute to information about the current speed and acceleration.
Well, on a straight section anyway, if you turn, the inner/outer edges will appear differently-spaced...
Wouldn't do anything positive in terms of suspicious staff though. They might think it's like some kind of heist-movie that will involve a specialized glass cutter making a circle on the window.
and then paint certain ties white and other black, now you have the world's biggest barcode
Why not just use phones GPS?
In the presentation they mention they are using GPS, and also an accelerometer.
In a Faraday cage?
with any view of the sky multi-band gps is damn good these days, my watch gets good locks through airplane windows even in the middle seat. In aluminum 737s, not compostite frames.
Train carriages, especially older ones, often have metallic windows so they are even worse than a typical plane. Even right up against the window, with the phone directly against it I have failed to get any lock on some trains.
Basically the inverse of https://daniel.lawrence.lu/photos/ or https://trains.jo-m.ch/ ;)
Also reminds me of the work by Adam Magyar, high-speed filming of people as the train arrives:
The former is in the prior art section at 2:24!
I've played with panorama mode and tried both of these (1 - panorama of moving train passing platform, 2 - panorama out window of a moving train)
...and both got the same weird artifacts, as it piled "similar" slices together then jumbled up "complex" slices.
Too bad they can't have a toggle for time-slice based panorama vs scene-based panorama. or maybe turn on accelerometer optimizations or not.
Free slit scan photograph apps exist
One fun thing is that the camera is that the projection is perspective in the vertical axis and othographic in the horizontal axis. That's why "zooming out" is done by squashing the image horizontally.
Through a regular camera, a person near the track, and a statue 16x times as large but 16x further away would have the same apparent size. They would show up as the same number of pixels in width and height. However, with a line scan camera, the apparent width of objects does not change with distance. Those same objects would have the same apparent height, but the width of the larger statue would be 16x.
You can also think of the camera of having a virtual sensor that's a few cm tall and a variable width, from meters to miles wide. It can only capture scenes as wide as the virtual sensor, unless it's being rotated.
Random thought: if you stick a camera to each train you might know their location better than Deutsche Bahn does.
Since at least you know where it was when you put the camera on it?
I love this so much!
This is such a cool hack, just for the sake of it. Amazing hacking and amazing art!
In the preso, she has said the full size output is on the website-
can anyone see what that link is? I didn't find it on the video page, but I might've just missed it.
http://philo.gay/linecam/gallery.html -- it's linked within the middle of the talk.
When I read the title I thought of another IDEA that I've had bumping around in my head that would be pretty cool.
Put down two model railroad straightaways next to each other, with two locomotives, use them as a base to attach a 3d-printer arm across and "print" a model railroad diorama, incrementally moving the locomotives to print each slice.
In other words, a flatbed 3d printer where one axis is the locomotives moving across your railroad layout. I think it could rejuvenate model railroads if you could print mountains, towns, streambeds, roads, etc...
A big part of the joy of model railroading is the building of such things, at least for some people.¹ It’s less about the product than the process. That’s a big part of why things like kitbashing and scratch building are popular. Finding a way to construct something out of nothing.
⸻
1. Yes, there are people who are more into the operations side of things, but I think for most model railroaders it’s as much about the building as the running.
You could take this as far as you want.
Like you could put down the foundation of everything, then manually add the grass or the buildings or pour epoxy water.
or you could print the roadbeds and the buildings and just paint everything yourself.
I wonder if you could create those "standard sized diorama" things that people take to railroad conventions that people hook together with their neighbors who also have standardized layouts.
You need a fairly precise movement mechanism for 3D printing.
Good thing we have those!
In your model railway? It can move the trains with 0.05mm precision?
From experience: the iPhone’s panorama photo mode works to take pictures like this from a train.
Using the camera as a scanner has a long history in satellite cameras and other aerial photography
This is often called Push broom scanning.
they need to connect with the skunkworks team, they solved this problem on sr 71 long time back !
That plane still looks bad-ass today. I heard that they just leak oil constantly on the ground before the engines warm up and the metal expands the the O rings make a seal.
well, they used film, which made it very easy to implement.
this is easy with a roll of film and a slit for an exposure window. you have the lens and the aperture and focus and all of that, but instead of a full frame, you mask it down to a small slit. slowly pull the film across the slit when it's time to capture your target. captures are limited only by how much film you can fit on a reel and feed through the camera during flight. I would expect that miles of film could be carried on the SR-71 without problems. You'd need (I assume) some custom machines to develop that film properly and to produce prints, but that's not a challenge for an entity that can design and build an SR-71.
It would have been pretty easy to use a horizontal line scanner to get image speed.
In any case, it seems that using a full frame video would be much easier to post process since frame to frame correlation tells you everything you need to know about motion.
One of the first scripts I vibecoded did exactly this. It works well, but is limited by framerate when there's high motion.
The first attempt was with video, but she later switched to exactly that, a 19000 lines per second 2048 pixel line scan imager. It was augmented with 6dof acceleration/gyro sensor, gps, and microcontroller to get the data to a laptop.
I'm not sure that video would have been easier to postprocess. At what fps and shutter speed?