Using my camera as a speed gun

[jaypatel33]

Hi,

Am brand new to this forum. I have recorded my bowling action (the english game of cricket, where you throw/bowl the ball the the batsmen). It was recorded side on to my run up. I wanted to measure my speed, so I thought, using 2 consecutive frames from movie maker, giving a time of 0.04 sec. Then using Pixlr(free photoshop), overlap the two frames, and measure the pixel distance the ball has traveled. Then I use the diameter of the ball (known) to measure how much distance each pixel is in real life. Then I calculate speed.

This should be accurate, with the only uncertainty coming from the pixel measurements. The results so far have me questioning if there are other factors I have missed, I got most of readings at 40mph, this is very slow, spinners bowl at this speed, I have been told I am faster, then I have one reading at 75mph. The fast one showed really good technique, so is why is that fast, but the slower ones shouldn't be that slow.

Thanks so much for the help.

Jay

[aedipuss]

you might get better results using a longer known distance. like count the number of frames from the ball leaving your hand until it crosses the home plate. there must be set distance for that if it's like american baseball.

[Cornucopia]

So what you're guessing is that if a C.B. is ~3.5cm and onscreen that 3.5cm take up 35pixels, then each pixel represents 1mm of width. And if each field (if interlaced) or frame (if progressive) is 1/50th of a second, that's 0.02 seconds per frame. So if your ball travels say 600 pixels (comparing leading edge to leading edge), that is 600/0.02 = 30000mm/sec = 30m/sec = 67mph. Right?

Note that in this instance your ball might be just barely showing on the left hand side of the screen in one frame (say 10 pixels), but in the next frame you would have the ball showing at 610pixels. And the largest (aka fastest) difference you can get is 718 pixels where 1 pixel of original leading edge + 718 pixels of movement + 1pixel of ending leading edge gap uses up the full 720 pixels (assuming SD resolution). So for this SD camera, the fastest that can be recorded with this method is 80.31mph. An HD camera, being widescreen and of higher resolution, would be of better benefit here.

So, you need to make sure your calculations are correct. Include units to assist in the conversion.

Note that as your distance away from the line of the ball's travel increases, the amount of ERROR increases, but zooming/moving closer decreases the available top speed. (I'm guessing that error is part of the inconsistency you encountered).

This is why the pros either use a real speed gun, or if they are actually trying to use the video footage, they might use a setup that includes a SERIES of finely-measured evenly-spaced, fully time-synchronized cameras, using the highest framerate possible.

Scott

[Cornucopia]

you might get better results using a longer known distance. like count the number of frames from the ball leaving your hand until it crosses the home plate. there must be set distance for that if it's like american baseball.

Yes, that's a possibility also (yes, there is a set distance in cricket), though the greater distance overall makes that error (+ or - margin) larger like I mentioned. But note you will then get only an AVERAGE speed for the whole travel, not any "instantaneous" speeds.

Scott

[jaypatel33]

So what you're guessing is that if a C.B. is ~3.5cm and onscreen that 3.5cm take up 35pixels, then each pixel represents 1mm of width. And if each field (if interlaced) or frame (if progressive) is 1/50th of a second, that's 0.02 seconds per frame. So if your ball travels say 600 pixels (comparing leading edge to leading edge), that is 600/0.02 = 30000mm/sec = 30m/sec = 67mph. Right?

Note that in this instance your ball might be just barely showing on the left hand side of the screen in one frame (say 10 pixels), but in the next frame you would have the ball showing at 610pixels. And the largest (aka fastest) difference you can get is 718 pixels where 1 pixel of original leading edge + 718 pixels of movement + 1pixel of ending leading edge gap uses up the full 720 pixels (assuming SD resolution). So for this SD camera, the fastest that can be recorded with this method is 80.31mph. An HD camera, being widescreen and of higher resolution, would be of better benefit here.

So, you need to make sure your calculations are correct. Include units to assist in the conversion.

Note that as your distance away from the line of the ball's travel increases, the amount of ERROR increases, but zooming/moving closer decreases the available top speed. (I'm guessing that error is part of the inconsistency you encountered).

This is why the pros either use a real speed gun, or if they are actually trying to use the video footage, they might use a setup that includes a SERIES of finely-measured evenly-spaced, fully time-synchronized cameras, using the highest framerate possible.

Scott

Hi thanks for the reply, yea the way you said it was correct, although the frame rate is 0.04s. Also in rough terms when I open the overlapped image on paint, the ball is around 6 pixels and the distance the ball traveled is around 100 (for the fastest ball), around 50-60 for the rest. I checked and re-checked the units.

[jaypatel33]

you might get better results using a longer known distance. like count the number of frames from the ball leaving your hand until it crosses the home plate. there must be set distance for that if it's like american baseball.

Yea, there is a set distance, unlike baseball however, the ball usually bounces before it hits the wicket (homeplate). I wanted to get a out the hand measurement because that way you don't have to account for the air resistance and bounce, also the I wanted to compare to the pros, whos speed is measured out the hand.

Thanks,

[aedipuss]

was the camera fixed on a tripod?

[Cornucopia]

If you are shooting PAL, the framerate is either 25FPS interlaced (SD, or 1080i) or 50FPS progressive (720p, 1080p). If interlaced, you should be using the FIELDS, not the FRAMES. This means the change between fields (at 25i) or frames (at 50p) is 1/50th sec or 0.02 sec between changes. That is a finer standard with which to calculate than using Frames on Interlaced. I'd use a BOB deinterlacer to get those frames out of the fields.

Scott

[jaypatel33]

was the camera fixed on a tripod?

NO, but was stationary on the ground

[aedipuss]

i would think the video of the ball would be a bit fuzzy and possibly elongated from fairly slow shutter speeds making distance as determined from ball size a problem. next time you might mark with chalk a known distance of a meter or 2 on the ground if you can put the cam on a tripod or put up a 1 meter wide piece of board beside the throwing lane that the ball would pass in front of and count frames.

[jaypatel33]

If you are shooting PAL, the framerate is either 25FPS interlaced (SD, or 10808i) or 50FPS progressive (720p, 1080p). If interlaced, you should be using the FIELDS, not the FRAMES. This means the change between fields (at 25i) or frames (at 50p) is 1/50th sec or 0.02 sec between changes. That is a finer standard with which to calucalte than using Frames on Interlaced. I'd use a BOB deinterlacer to get those frames out of the fields.

Scott

Wow, there was a lot of stuff there that I have never heard of. I just did right click, properties, details and it said 30fps. I'll google some of the phrases you used and will reply soon. Thanks

[jaypatel33]

i would think the video of the ball would be a bit fuzzy and possibly elongated from fairly slow shutter speeds making distance as determined from ball size a problem. next time you might mark with chalk a known distance of a meter or 2 on the ground if you can put the cam on a tripod or put up a 1 meter wide piece of board beside the throwing lane that the ball would pass in front of and count frames.

Might give that a go, thank. although the ball does seem pretty clear, and spherical, and I sharpen it when editing.

[Cornucopia]

You also want to use a cam that has an adjustable shutter, and make it as short/fast as possible (e.g.: 1/10000 sec). Then, you won't get the "fuzzy" problem mentioned by aedipuss (though you will need lots more light, but that shouldn't be a problem outside in the daytime).

Scott

[budwzr]

Put a tape measure in the background.

[Cornucopia]

Put a tape measure in the background.

No that won't work right, unless it is IMMEDIATELY behind/beside the path of the ball's travel (due to perspective).

Scott

[budwzr]

Put a tape measure in the background.

No that won't work right, unless it is IMMEDIATELY behind/beside the path of the ball's travel (due to perspective).

Scott

Oh yeah, I forgot.

What about duct tape with ticks on it? Nah, won't stick to grass. Hang on, I'm thinkin".

Got it! Somebody shoots a cannon ball at a known velocity and extracts the unknown velocity from the known one. Prollem wit dat doe, is how do you syncronize the camera with the cannon?

Another option, but this is expensive, is to go ahead and kick out the $29 and buy one of those golfer gadgets that also has a can opener, wind speed, weather, etc.

[jaypatel33]

You also want to use a cam that has an adjustable shutter, and make it as short/fast as possible (e.g.: 1/10000 sec). Then, you won't get the "fuzzy" problem mentioned by aedipuss (though you will need lots more light, but that shouldn't be a problem outside in the daytime).

Scott

Ok, so I kinda understand progressive interlaced, I'm using a digital camera video recorder, so that would be progressive, so I method is right, using 2 frames, right?? Also, why are you mentioned shutter, is that not for photos, I;m using a video recorder, maybe I was being unclear.

[jagabo]

Measure the distance traveled over several frames, not just one. That will give you a more accurate measurement. A 1 or 2 pixels measurement error over 10 pixels of movement (one frame) is a 10 to 20 percent error. That same 1 or 2 pixels over 100 pixels (10 frames) is a 1 or 2 percent error.

If your camera is at a fixed location you can use a more accurate measurement than the size of the ball. Before shooting have someone hold up a meter stick (or tape measure) at the plane of interest.

If the camera has a zoom lens, put it far away and use the zoom. That will reduce fisheye type distortion.

As was already mentioned, use a fast shutter speed to get a clearer image of the ball. If the camera doesn't have manual control use and "action" or "sports" preset.

[Cornucopia]

@jagabo, if the ball is traveling as fast as I would think it would travel (~60-80 mph), you won't get 2 full, clean frames with which to calculate (as the ball will be partly or fully out of frame in one of the succeeding frames), unless you zoom out, in which case you will be increasing the margin of error anyway.

@jaypatell33, the SHUTTER that jagabo and I mentioned is inherent in ALL cameras, whether still or motion (some are mechanical, some electronic). This SS is mainly independent of the framerate in video/motion cameras. On cheaper cameras, the FR and the Shutterspeed are exactly or almost the same. IOW, the shutter remains OPEN (allowing light in) for the duration of the framerate. On better cameras, this can be shortened at your discretion.

Example Of Frame = Shutter (where High = letting light in, Low = keeping light out):

__|----|__|----|__|----|__

Example of much shorter/faster shutter:

____|-|____|-|____|-|__

edit: ok stupid text spacing didn't line up like I hoped!

If you look closely, you'll see that the RATE that each frame cycles in time is the same, but the amount of time within each frame where light is allowed in is less. Less accumulated light = dimmer, but shorter time = sharper image.

Also, your mention of camera's having only progressive recording is a shaky assumption. Even with DSLRs, I don't think it's a solid rule. Check your manual. If it says 50i, you can bob deinterlace and use the frame from each field to work with a sharper image. If it's 50p, you're as already good as you can get (for consumers). If you truly have a 25p, your motion resolution is more limited and there is nothing you can do about it. In which case, you were correct about the 0.04 sec framerate.

Scott