Epic Motion Sensor & GPS

[Shane Betts]

After looking at that 30th Oct thread, with over 1,400 replies and 3/4 million views, I thought it might be good to start a couple of threads specific to individual features of the new cameras, just to reduce the amount of trawling required, provided you know what you're interested in researching. So here goes.

[Shane Betts]

I'm amazed at how much thought has gone into the DSMC line and this is a classic example of Red going way beyond the call of duty.

So, what are the possible ramifications/applications of the motion sensors? Can they be used to govern the way the camera moves (by communicating with stability devices for instance) or do they simply log metadata to aid in compositing in post?

[Philip Hodgetts]

Well, I'm guessing that if you tracked the output of the motion sensors you could use the metadata to input to a 3D camera in post/scene rendering. And the GPS would fit right into a workflow that used Derived Metadata. (My term in trying to build a taxonomy of metadata http://www.philiphodgetts.com/2009/01/30/what-are-the-different-types-of-metadata-we-can-use-in-production-and-post-production/) There are actually six types.

I'm excited to get GPS metadata from a camera to use in our editing algorithms too.

Philip

[Shane Betts]

Can't wait to see what you do with it Phillip. Makes me laugh when people try to compare the Epic with cameras like the 7D. One is an elevator and the other is a length of rope. Yeah, they might both get you to your floor but...

[Kevin Olsen]

GPS at this level will only be accurate to about 7-10 meters. I don't know exactly how helpful that is, but maybe it could be.

Seems like the motion sensing could be used in combination with fly-by-wire to deliver constant focus while moving the camera around the object.

E.G. you could mimic the 'Dolly Zoom' by using the motion sensor to detect backwards movement and tell the fly-by-wire system to zoom in and maintain focus on the specified target. This could apply to many different moves that would require less hand work and just a bit of calculation.

KO

[David Wyatt]

This may well be "one of those stories" :001_rolleyes: but I heard George Lucas was looking into using military-grade GPS (more accurate than normal GPS) for CG purposes before embarking on "Phantom Menace". Needless to say it didn't work out (probably for matters of "National Security" :patriot:)

[Shane Betts]

I have no idea what I'm talking about here but I'm sure someone on here will know more.

Global Positioning is designed to establish your position in reference to the globe, by trilateration from satellites orbiting the Earth and transmitting signals for that very purpose. And, due to the enormous distances involved, those calculations have a limit of accuracy because there are things between the signal source (satellite) and the receiver (in this case the camera) that can subtly alter the speed of the transmission from speed of light - right?

Any GPS device is designed to receive signals only from GPS satellites, make the calculations and then disseminate the data. But what if a GPS device had a secondary (alternative) setting where it would receive only the signals from, say, four mated devices transmitting from close by (say, within 20m). Those transmitters would need only to transmit a very small signal (think Bluetooth) and the GPS/LPS system would then be able to calculate within fractions of millimetres. Of course, it wouldn't be able to tell you how far you were from the nearest town but it would provide valuable information on location and angle of the camera in comparison to a reference point - greatly assisting in post as well as, possibly, camera stabilisation. Is this possible?

[Jim Collins]

The speed of light in a vacuum is a constant and the speed of light can't be higher than that. Any time light is passing through any sort of material, however, the speed will be lower and can be a lot lower.

This is very important because this is how a lens works. The glass slows down the light so the light is slowed down for a longer time when the light is gong through the thick part of the lens than when it is going through the thin part of the lens. For a converging lens where the glass is thickest in the middle the light is slowed down more in the middle than the edges so the light is bent inward. High index glass is glass that has a lower speed of light which allows the lens to be thinner for the same affect.

This is important for GPS because discontinuities in the atmosphere will affect signals from different satellites in different and changing ways which reduces the accuracy of the position. Military receivers, and civilian receivers soon too, use GPS signals on different frequencies to measure these effects and compensate for them.

[Shane Betts]

And, of course, it isn't light that is being captured by a GPS receiver, it's a radio frequency. My understanding of it (as I have mentioned, it's limited) is that the satellites broadcast 50 bursts per second. Not sure if that cadence also has some bearing on the accuracy or not as well. Nonetheless, my point remains. A custom set up, local grid would provide all the accuracy we need. In a feature film environment this would seem extremely doable. Lasers of course would cause all sorts of problems in a film set. The actors would start speaking in Austrian accents for a start.

[Noah Kadner]

Yeah I'd guess something far more accurate than GPS would be needed for data useful to 3D tracking...

Noah

[Peter Mosiman]

I think GPS will also help in terms of tracking down stolen cameras.

There was a lot of talk awhile back about installing GPS into camera's to prevent theft of something so expensive. Maybe these chips have multiple uses? (metadata + theft prevention?)

[Noah Kadner]

I think GPS will also help in terms of tracking down stolen cameras.

There was a lot of talk awhile back about installing GPS into camera's to prevent theft of something so expensive. Maybe these chips have multiple uses? (metadata + theft prevention?)

Now that's what oughta be a no-brainer considering there's a GPS chip in every $100 cell phone these days.

Noah

[Andrew Ravani]

Think about GPS as an approximate location guide. Now think about how bad AND varied the offset is on GPS.

Now think about the fact that you really don't care - because in a two or more camera production, you are both going to be receiving the same offset information.

Now add in the inclinometer and you have two very interesting tools for spatial coordination of cameras - a way to reconstruct camera movement well after the fact.

Now add i/lens data and it all starts to make a lot of sense - that coupled with the new electronics packages on the RED lenses set to come out...

Now if you are a production lucky enough to have two or more Epics with RED lenses, you sync time code, then record all of your meta data and make a few simple measurements and do a "Spatial slate" - shooting a common element in space and time - then you can pretty accurately reconstruct the spatial environment based on lens info, GPS spatial relationship between cameras, and individual relative motion of the cameras ... holy shit.

[Jason Sinclair]

Can someone explain: if you have bad data on the gps, exactly how can you make this into precision data with two lots of bad data? Can someone point me to a white paper or similar?

[Eric Haase]

I think they put GPS in there so you can suction cup your Epic and touchscreen LCD to your windshield and get turn by turn on your way home from the shoot.

[Deanan]

I think they put GPS in there so you can suction cup your Epic and touchscreen LCD to your windshield and get turn by turn on your way home from the shoot.

It's so you don't get lost on the way to the shoot at 5am.

[James Brundige]

gps is dithered, except for the military, and can't be accurate to less than 10 meters or more.

but the accelerometer is awesome. could read the metadata and completely eliminate camera shake. (I never shake, but my airplane does)

[Harry Lipnick]

gps is dithered, except for the military, and can't be accurate to less than 10 meters or more.

but the accelerometer is awesome. could read the metadata and completely eliminate camera shake. (I never shake, but my airplane does)

Actually, with the addition of ground based augmentation, called WAAS, GPS can be slightly more accurate (7m theoretically, but often even better in practice.) http://en.wikipedia.org/wiki/Wide_Area_Augmentation_System#Accuracy
Right now, this is being used for precision aircraft approaches in both military and civilian flying. Of course I have no idea if the RED cameras would be able to tap into WAAS, nor if RED's plan for GPS would require this kind of accuracy. Either way, it's incredibly cool technology.
Peace,

-Harry

[Kevin Olsen]

Civilian systems exist up to accuracy of 1cm but are only commonly used for surveying and require a still receiver and plenty of time to calculate location. This level of system is a Differential GPS system and accomplishes it's calculation by transmitting the error from a known Geodetic landmark and 'subtracting' the error between the satellite and landmark signals. Think of this like you would balanced audio, where the difference in signals is subtracted and noise is removed, only this case it's time error.

The real problem with GPS is it requires line of sight to a minimum of four satellites to obtain accurate locations. This often rules out any indoor scenarios and those outdoor locations with many obstructions nearby.

Start moving the receiver and your screwed.

KO

[Jason Sinclair]

Im glad we cleared that all up.

[Illya Friedman]

Yeah I'd guess something far more accurate than GPS would be needed for data useful to 3D tracking...

Noah

Not necessarily. I've seen a demonstration of milspec GPS accurate to approximately 1 cm (in x, y & z). Understandably, it was very (very) pricey.

If calibrated to the cameras nodal it could be used for painless 3D/Virtual Sets, and a host of other related applications.

I encourage any/every camera manufacturer to invest in this concept, as it would make high quality VFX post production a breeze. To make it truly useful, the coordinates should be stored in the individual frame metadata.

I know this kind of GPS integration can (and will) be done on a mass scale at some point, I just don't know when that will be. The future is bright for this technology.

I.

[jbeale]

GPS has come a long way since the 1970s, both satellites and receivers, and secondary correction systems eg. DGPS. Going further and using "pseudolites" (placing your own local GPS transmitters at known fixed coordinates) you can eliminate all the complications with ionosphere delays and get real-time precision at the 1 cm level over a substantial local area (kilometers). Not sure if this could work with a typical consumer GPS receiver, and if it did, I'd assume you would need specialized firmware on it.

Stewart Cobb did his thesis on this at Stanford in 1997: http://waas.stanford.edu/~wwu/papers/gps/PDF/stuthesis.pdf

Edit: after some more reading, it turns out the research in the thesis above led to a startup company, which is now called Novariant, and they sell commercial "sub-inch" class positioning systems, based on GPS augmented by pseudolites. Main application right now seems to be automated farming; eg. tractors that plow very accurate furrows. See also: http://www.novariant.com/

[Jason Sinclair]

1cm inaccuracy is not good enough. There are other methods worth trying.

[JohnF]

This may not have anything to do with tracking features but an in-camera GPS could also provide a very accurate clock/timecode source.

If such a feature is enabled one should be able to tell all cameras to read GPS time as timecode by default. Very handy for a multiple camera shoot.

Furthermore GPS time is so accurate that it may be possible to actually sync cameras using it. Even handier for multiple camera shoot.

JohnF

[Dave Neathery]

This may not have anything to do with tracking features but an in-camera GPS could also provide a very accurate clock/timecode source.

If such a feature is enabled one should be able to tell all cameras to read GPS time as timecode by default. Very handy for a multiple camera shoot.

Furthermore GPS time is so accurate that it may be possible to actually sync cameras using it. Even handier for multiple camera shoot.

JohnF

I keep hoping they will do this. Right now, I have to run two coaxs to each camera: One for TriLevel sync and one to loop the timecode from cam to cam.
(It works perfectly, but it can be more than a little inconvenient at times.)

Dave

[Nils Ruinet]

This may not have anything to do with tracking features but an in-camera GPS could also provide a very accurate clock/timecode source.

If such a feature is enabled one should be able to tell all cameras to read GPS time as timecode by default. Very handy for a multiple camera shoot.

Furthermore GPS time is so accurate that it may be possible to actually sync cameras using it. Even handier for multiple camera shoot.

JohnF
This is definately what I'm hoping for...