Filming a passing road sign: speed vs bumpiness

[guyburns]

As part of an audio-visual I'm producing, I want to video two passing road signs while driving in a car on a country road. The videos have to fit in with commentary and music, so the timing is pretty tight, and I want to get it right the first time I visit the area – 4 hours drive from home. Each video will be about 5 seconds – road sign appears in the distance, and then we pass it. I can think of several methods, each with its pros and cons, and I'd like some input from people more technically minded than I am.

I can't be sure at what speed I should drive past the signs, because the video will be constrained by the music and commentary. It might appear on screen for 3 seconds, but it could be 7 seconds, so I thought what I should do is video at 20kph, and then I can make it appear faster, if needed, by dropping frames. For example, if every second frame is drop, the video should appear to have been taken at 40kph, and so on for other frame-drop scenarios.

Two questions arise. I'll use the word "bump" to mean vertical undulations, not actual pot holes and the like.

QUES 1. I assume that visually, on screen, there should be no difference between a video shot at, say, 80kph, versus one shot at 20kph but 3 in every 4 frames are dropped. Is that the case?

QUES 2. Bumpiness. Will shooting at 20kph decrease the bumpiness of the video? Or increase it? If you drive slow enough, the car will follow every single bump, up and down, whereas at a faster speed there may be a smoothing effect because the car rides over the bumps. But a faster speed may cause overshoot in vertical movement in some (all?) cases.

Any suggestions welcomed, especially about the relative bumpiness of video shot at 20kph and speeded up, versus being shot as 80kph.

A third possibility
I walk on the edge of the road, taking a still photograph every step (about 0.7m for me), aimed at the middle of the sign. Then I turn those stills into a video. The AV will eventually be shown at 24 frames per second, so the stills approach will result in a car speed of ~16.8 mps, almost exactly 60 kph.

This method has the advantage that there will be no bumpiness caused by the road surface, and it will have a distinctive look because as I walk past the sign taking stills, on screen it will appear that the camera has turned almost 180 degrees – unusual for a video shot from a car. Normally video cameras on cars are fixed in place. Problem is: I'll have to take a couple of hundred stills.

Any comments on the effectiveness of this method?

I will be testing these methods locally, but thought I should get some physics input beforehand.

 

[Drakkith]

QUES 1. I assume that visually, on screen, there should be no difference between a video shot at, say, 80kph, versus one shot at 20kph but 3 in every 4 frames are dropped. Is that the case?

There will be a difference in any background motion that gets captured. You're essentially speeding the 20 kph film up by dropping frames, so any background (or even foreground) motion will appear faster than normal. This may or may not be detrimental to your shot. The reverse is true if you take the 80 kph shot and slow it down by cloning frames or interpolating between frames, where you will slow down any motion in the shot. Assuming there's no noticeable motion for anything besides the sign, then you should be okay slowing/speeding up the shot.

QUES 2. Bumpiness. Will shooting at 20kph decrease the bumpiness of the video? Or increase it? If you drive slow enough, the car will follow every single bump, up and down, whereas at a faster speed there may be a smoothing effect because the car rides over the bumps. But a faster speed may cause overshoot in vertical movement in some (all?) cases.

This is difficult to answer because different types of dips, bumps, potholes, etc all affect the vehicle differently at different speeds. We could talk theory all day but in the end your best bet is to simply take the shot at a certain speed and see how it turns out.

A third possibility
I walk on the edge of the road, taking a still photograph every step (about 0.7m for me), aimed at the middle of the sign. Then I turn those stills into a video. The AV will eventually be shown at 24 frames per second, so the stills approach will result in a car speed of ~16.8 mps, almost exactly 60 kph.

This method has the advantage that there will be no bumpiness caused by the road surface, and it will have a distinctive look because as I walk past the sign taking stills, on screen it will appear that the camera has turned almost 180 degrees – unusual for a video shot from a car. Normally video cameras on cars are fixed in place. Problem is: I'll have to take a couple of hundred stills.

Any comments on the effectiveness of this method?

That sounds like a perfectly good method too. Really it boils down to how you want your shot to look.

Honestly, this appears to be a relatively simple shot and I'd bet you could do all three methods in a relatively short period of time. If your camera (for stills) supports automatically taking shots every X amount of time then even the third method should take only a few minutes at most, as you can just walk forward at whatever speed you want while the camera automatically takes pictures.

 

[Baluncore]

What is the road surface material like? Could you drive along the crest of the road, rather than in the ruts made by regular traffic?

Go to Google Earth and use Street View to look at the road surface and the sign.
Google will probably automatically have blurred the image of the sign.

Give us the Lat, Long and we can look at the set.

 

[Ibix]

You can certainly knock out frames, but I'd suspect that messing with the frame rate to the extent of one frame in two is going to make things look a bit weird, particularly if there are curves in the road. The way you take a bend at 80mph is very different from the way you take it at 40mph and it'll show. 50mph vs 40mph, less so.

Do you know how far apart the signs are? Google Maps and Streetview will help you there. Then read the script and time it, so you know how long you need between the signs. From that you can work out the speed you need to be going. Film while driving at slightly below that speed and do some fine tuning by dropping a few frames when you have the final footage and the final commentary.

Also, and I cannot stress this enough, experiment. You need those particular signs for your shot, I presume, but there are loads of road signs all over the country. Find some random pair in your neighbourhood about the right distance apart and film them, and try your techniques on your test films.

 

[sophiecentaur]

As part of an audio-visual I'm producing, I want to video two passing road signs while driving in a car on a country road.

How 'smooth' do you want the sequence to be? The resulting video will depend as much on the suspension characteristics of the car as the road surface. Shooting from a luxury car will produce an entirely diffrent result than from an unladen truck with a harde suspension. Film makers go to a lot of trouble with camera suspension if they want a very smooth result. Pitch and roll can have as great an effect as vertical displacement and also the position of the camera in the vehicle.
Daft question, maybe but why does it matter? Are you trying to get some data from the movie or some subjective effect.

Stop motion photography could be very hard work because you will need some sophisticated mounting for the camera. What would you point it at? In fact, what would you point it at when shooting the movie? At the horizon, at a mid field object or kust keepong in a constant compass bearing (also a bit difficult.

Why don't you just try things out on a local road with any old sign before you go to your location? Don't drive four hours to test your ideas.

PS there are many frame rate converters available - some actually free. They are used for slomo in sport all the time.

 

[guyburns]

Thanks for the responses, they've given me something to think about. It looks like videoing at low speed will not decrease the bumpiness because of the variables involved. I had thought of letting air out of the tyres to make for a smoother ride, but from what respondents have said, that would just introduce another unquantifiable variable, so I've tossed that idea.

1. Drakkith: Taking photos automatically while walking is a possibility. My Lumix GH4 can do that, but there might be more shake from my walking motion than from inside a car. I'll give it a go though.
2. Baluncore: Road surface is sealed, two-lane, at 42ºS, 148º 04' 04"E. It's a country area, a few farms but not much else. I don't reckon there would be a Google street view.
3. Ibix: There is a long sweeping curve in the road just where one of the signs is. I hadn't thought of a curve interfering with the look of a speed-altered video. You could be right.
4. sophiecentaur: Yes, it probably doesn't matter if there is shake in the video. This is partly a technical exercise, trying to get as smooth a result as possible, for the learning experience.

I reckon I'll try two techniques:

1. A hand-held video from inside the car, slower than normal speed, keeping the camera aimed at the same angle in the general direction of the sign, capturing what a driver normally sees.
2. Take a hundred or so stills, each one aimed straight at the sign, even after I've walked past it. The will give the effect of the camera rapidly rotating to keep the sign in view as the sign moves away behind. I've never seen that effect in movies. It could be interesting.

Anyway, thanks for the input.

I'll do a few test runs locally, look at the results, and then head off to 42ºS. If I get a pleasing result, I post a download link so respondents can check out the most aesthetic sign-passing video ever made!

 

[Baluncore]

Road surface is sealed, two-lane, at 42ºS, 148º 04' 04"E. It's a country area, a few farms but not much else. I don't reckon there would be a Google street view.

I'm surprised, I am also under two hours from there, but only two minutes by Google Earth. Here is a street view from Google Earth.

 

[Vanadium 50]

Do you need to video it at all? Can you pan and zoom a still shot?

 

[Mister T]

I would drive by slowly, then edit the video to speed it up as necessary. This will smooth things out. There is also smoothing software IIRC.

It's pretty common to use video analysis to do introductory physics activities, so there's lots of software out there along with activities posted on the web.

 

[Baluncore]

The road there is narrow, with a couple of distinct glazed 'ruts' made by heavy vehicles. If you cruise through, with two wheels on the white line, and the other two between the ruts, you should get a steady recording.

A bigger problem will be matching the weather and shadows.

I can't be sure at what speed I should drive past the signs, because the video will be constrained by the music and commentary. It might appear on screen for 3 seconds, but it could be 7 seconds, so I thought what I should do is video at 20kph, and then I can make it appear faster, if needed, by dropping frames.

The timing can be done from Google Earth, Street View. Drop markers at the limits of sign visibility, measure the distance between markers (max≈250 m) and compute the equivalent speed required to fit the available time.
3 sec = 83.3 m/s = 300 kph.
4 sec = 62.5 m/s = 225 kph.
5 sec = 50.0 m/s = 190 kph.
6 sec = 41.7 m/s = 150 kph.
7 sec = 35.7 m/s = 129 kph.
The speed limit there is probably 100 kph, so you will need to reduce the travel distance while recording by a factor of two.

 

[Vanadium 50]

Do you need a video at all? Why not a still that you pan and zoom?

 

[pbuk]

It's 2024. Using a GoPro (other action cameras are available) will eliminate almost any uneven motion, and modern video editing software will slow down or speed up the clip by any amount you want smoothly, especially if you film at 4x.

 

[sophiecentaur]

Do you need a video at all? Why not a still that you pan and zoom?

You will lose the parallax effect, relative to the background objects. This may or may not be relevant to what the film maker wants but parallax says a lot about speed and motion.
The poor man's version could be to use the same image for both situations.

 

[Baluncore]

The poor man's version could be to use the same image for both situations.

The actual sign is shown in post #7. It is not a single flat sign, as it has two sides, which makes panning somewhat difficult, especially against the distant landscape background.

 

[sophiecentaur]

The actual sign is shown in post #7. It is not a single flat sign, as it has two sides, which makes panning somewhat difficult, especially against the distant landscape background.

That was my extreme (poor man's) solution.
My point was that the video script could require anything. PF is hardly qualified to do more than the basic Physics, which is a small part of the job. We don't actually know what's in the director's head. Short of pointing out that parallax is very relevant if you wan't to emphasise motion, there's little else we can advise on. The GoPro (etc.) technology does most of the thinking for you when it comes to image stabilising.

 

[guyburns]

Thanks for all the suggestions, especially to Baluncore for finding that sign on Google Earth. That's the one, and there's another a bit further along the road.

I went down there a few weeks ago, and yes, as per Baluncore, the road does have ruts and is rather bumpy. I made several passes, including a "stills" method, where I walked along the road taking 120 photos (5 seconds worth of video) aimed directly at the sign.

Well the stills idea was a failure. In addition to Premiere being unable to "Warp Stabilize" the stills into a smooth video, the other problem was the clouds sweeping across the sky very quickly. It took me about 20 minutes to take the 120 photos, so the clouds were being speeded up by a factor of 240 (20 x 60/5).

The Premiere failure I might ask about in the Premiere forum. I think it couldn't stabilize the stills successfully because, even though I tried to aim for the centre of the sign each time, there were abrupt jumps in the position of the sign on the stills. A normal video at 24fps, wouldn't have those sorts of jumps, even under bumpy conditions. There might be a lot of shaking around, but not sudden jumps from one frame to the next.

Anyway, it was an interesting test, and I've put the stills idea to rest. The videos came out okay.

If anyone's interested in the stills version, you can download it from https://www.mediafire.com/file_premium/bpswruzhzglmp67/Royal_George_%28Stills_2k%29.mov/file

 

[sophiecentaur]

the other problem was the clouds sweeping across the sky very quickly.

HAHA. All those clever clogs PF readers and no one thought of that.
Glad the video worked out OK!