Video Metrics - Vid Analysis Data Points for PV Performances

[master]

Most digital video cameras "record" in a relatively uncompressed format. It is what is done to them after that makes them .mov, .flv, .wmv, .rm, .mpg, .mp4, etc. And, in doing that conversion the operator of the software decides just how much file compression (read loss of detail for benefit of file size reduction) is applied. The point being not all .mov files are of equal quality unless they were processed with the same compression algorithm. That applies to any of the formats. This algorithm is commonly referred to as a CODEC (COmpress DECompress). Perhaps you have had the experience of trying to play a particular video in a player and it not working. The usual culprit is that it was compressed with a codec that you don't have installed for your player so it can be decompressed.

Someone above said the best quality is seen when played directly from the media as recorded. That is because it hasn't gone through any compression or conversion process.

For good quality stop motion video, your camera also needs to have at least a "sport" mode or better yet, an actual shutter speed setting. Of course to use a fast shutter speed, you need more light and/or a high quality image IC which has more signal response in low light conditions. How many videos have we all tried to look at that were shot in poor light conditions. Certain cameras are better than others with this regard. And all "consumer" cameras are limited to the 30fps limitation that was also described above.

It is an extremely difficult task to gather video from random sources and expect to be able to process it in a way as to make meaningful measurements of the resolution you desire for pole vault analysis. I'm pretty sure a lot of these limitations were discussed in previous posts/topics.

- master

[golfdane]

I have previously written about this piece of software: Skillspector/Skillcapture
It's FREE...... The backdrafts are:
Video capture requires a firewire attached videocamera (it will work with a USB webcam, but the quality will suffer greatly in terms of fps and resolution).
It cannot analyze any of the known formats of video (mov, mpg, most variants of avi (divx,xvid))

http://video4coach.com/index.php?lang=english

[dj]

hello

What is the plan?

We are simply looking at proportions and comparative analysis….

Whither it is the proportion of the number of frames or a proportion of the total time you can determine.. action reaction, application of force ..etc..

And I don’t know if this will help the “argument” but…….

For example when you realize that speed is a product of how much force is applied with the ‘downward” action of the leg.. and that “contact” time is the deciding fact because “air” time will be the same for everyone(provided they are not bouncing up and down) and that the “time and distance” of the “knee lift” to back to contact with the surface is THE most important factor, you can take any quality film and do a proportionate comparison by frame or time.. from one race to the next or from one athlete to the next..

So what I want to emphasize or add in this discussion is we can and should do all our final analysis by comparing “time and distance”. A “picture” does not tell the story.. comparative math and “time and distance” is always the final word… physics.

Why could Joe Dial or even Tim Mack seem to defy the “charts”.. your looking at the wrong charts.. your looking at pictures that indicate the “bumble bee can’t fly” but he can because of time and distance.. the bumble bee because he moved those stubby wings so fast he could get his fat butt into the air.. T-Mack and Joe because of the same application of force by time and distance…….

dj

[ADTF Academy]

Sony has a realitvely inexpensive high speed camcorder on the market. It is available in standard and high definition.

It takes the standard 60 fps and quads it to 240 fps over a 3 second window. This option is allowed to be recorded in the future or what happen 3 seconds in the past. IE a vaulter can go land on the mat and than you hit record. The camera records what it saw in slow motion three seconds prior to the landing. Very cool. Great for instant feedback as well cause you can watch it as it records to confirm what you saw.

I also have dartfish and have used the videos shot with this camera for the speed based training and development and plugged them into dartfish to get very acurate speed ratings. I know this because the same area I measured with the camera and dartfish I also timed using speed traps and the two figures where exactly the same.

Except for Peter micginnis at large meets i don't think anyone is allowed to place markers down on the track. I wish I would have asked at Pre last year and will do so at some of the larger meets we attend this summer to get video of most of the top guys in the US jumping. Simply put youtube vidoes are worth about as much as they cost to watch them.... nothing.

At the tirals I didnt' end up filming due to Mark not making the finals, but if anyone has a stantionary video of the meet I know Peter had 5 to 10 meter marks down on the runway. With these marks in place I can easily find the data your looking for in terms of distance covered and time it took to cover. Even the CDs we got from Peter from the Summit he doesn't hold steady he follows the vaulter down the runway and I feel makes even his speed figurs sightly scewed (i know he does the count the frame method). Plus it made my ability to use the footage obscelete for the purpose of using dartfish.

I will see if we can get some good video taken from Mt. Sac, Drake (mall and outdoor), Modesto and Pre plus Mark should be hopefully back jumping full again soon so I'll see what I can get from practice as well.

[rainbowgirl28]

Except for Peter micginnis at large meets i don't think anyone is allowed to place markers down on the track. I wish I would have asked at Pre last year and will do so at some of the larger meets we attend this summer to get video of most of the top guys in the US jumping. Simply put youtube vidoes are worth about as much as they cost to watch them.... nothing.

You can always try, but any meet that is going to be on TV probably won't like it much. Even Peter has a tough time with officials trying to move his marks.

[KirkB]

We need to be thinking "outside the box" more! Sometimes when we think about what we CAN'T do, we eliminate our options too early. We need to focus on what we CAN do! We need to be more innovative!

It would be great if we can scale to markers along the track, but if they're not there, all is not lost. Off the top of my head ... here's 3 other "markers" that I can think of to scale to ...

1. The pole length or pole grip whilst the vaulter is approaching his takeoff. Either one will do.

2. The vaulter's height.

3. The standards. You need to know the height of the bar, and you need to have the full height within the camera view. If you're really lucky, you might even be able to read off the foot markers on the standards. Remember that they're skewed to the plane that the vaulter travels, so a little trig would be required to scale from them. Pogo has these trigonometry formulae pinned indelibly in his brain!

4. The width of the sign that shows the crossbar height. Someone will need to physically measure it, but once measured, every one ... at every meet ... should be the same ... from the same manufacturer. And I'm assuming a 90 degree angle to the camera. If not square to the camera, then this one's out.

Once you get one of these measurements, you can compute the other 3. Or if you have 2 of them, you can use them both to verify the accuracy of the PV Performance data points.

You also need a fixed point near the takeoff area. The sign showing the crossbar height might work for that, if properly located. This point should not be too distant from the takeoff area. I sometimes use a point on an official's body. Assuming they're standing still during the vault ... how much do they move in less than a second ... especially when they're usually just standing there?

And why does the marker need to be permanently affixed to the runway? Why not have your vaulter place his pole of known length (a) on the top of the box ... straight up and down (vertical scale); (b) on the top of the box, lying straight back down the runway (horizontal scale); and (c) the next pole length down towards to start of the runway. With the cooperation of your vaulter, you can be in the stands with your camera on its tripod, and take a couple secs of vid for these 3 measurements ... even well before the competition starts! Once these "marks" are on the vid, they're there for the life of the meet (as long as you don't move your tripod).

And don't change your zoom setting!

You can also use (b) to compute the takeoff point ... provided it's not hidden behind the pit aprons. Even if it is, take a couple secs of vids with your vaulter posing with his ideal grip and takeoff point ... and use that as a comparison to his competition vault takeoffs.

Now if you want to get fancy, you can also ...

1. Mark your poles with tape ... say exactly 3 meters from the top of the pole to the top of the strip of tape. This marker will always be visible during takeoff despite the aprons blocking the view of the box and the butt of your pole. This is yet another handy "scaled" measurement.

2. Put additional "identifying marks" on your pole ... just below the 3 meter line ... visible from a vidcam ... to positively identify the length of the pole ... and possibly even its flex. With your consultation and endorsement (all you PVP contributors), we can propose a world-wide standard for this.

I'm serious! Even if we didn't need to know the pole length and flex (which we do) for our PV Video Metrics purposes, wouldn't it be nice to positively identify these 2 stats for ANY vaulter's vid that you watch? Just be "seeing it" on the vid? Once we develop this standard, we can encourage the manufacturers to apply the appropriate marks onto the poles at the factory. Why not?

Kirk

[dj]

good morning

great ideas Kirk.. i've done many of these.. i also put a stripe on the standards.. with white tape..

and all my poles have red painted "racing stripes" at measured distances.. not only do i know if someone has grabbed one of our poles i can get measurements..

dj

ps could you gather anything from the Tim/Daniel vid??

[Pogo Stick]

So here's my Elevator Pitch ...
[size=120]

[i]Pole Vault Video Metrics is a process of analyzing amateur PV videos - to help aspiring young vaulters compare the speeds of different parts of their vault, and to compare their speeds against elite vaulters. This will help them improve their technique, strength, speed, and ultimately their PR - by spotting their weaknesses, making improvements thru focussed training, and tracking their progress.

I really like your sales pitch. I would like to take step forward and try to define some technical requirements. Also we need Reference Point - a video(s) which can serve for testing basic premises and fine-tuning. All relevant parameters about this video (jump) need to be known and accurate. Any error in this process of adjusting will be carried forward. YouTube videos in most cases are not good enough. Even StabHochSprung QuickTime videos are often not so god. Just for illustration, here is a detail from Bubka's 601 video from StabHochSprung, enlarged 2x:


You can just guess where is upper hand.
Does someone have some suggestion?

I think camcorder itself is not an issue. Even if you go down to the very bottom end of MiniDV camcorders (Canon ZR900, Sony DCRHC52 or Samsung SC-D382) you can still have much better quality than above video. $190 (Canadians) is not too much for camcorder - this is half price of one pole.
As master said post-processing video is the main problem. Video format, codec and compression ratios have huge influence on end result.
Here are few suggestions:
- any miniDV camera is good. MiniDV videos are not so compressed than from memory card, DVD or HDD camcorders. We can test other cameras later
- find some free editing software for capturing and video conversion
- pick codec and video file format
- define video parameters like quality, compression and output resolution. Preserve original resolution and screen ratio
- standardize camera position, shooter speed and view field (both horizontal and vertical). Keep in mind that other countries (actually majority of the world population) use PAL/SECAM standard which has different resolution and frame rate
- if possible mark position of back of the box and make it visible on video (with tape over the pit or some other marker)
- decide what to measure and how. Reference video is the key

Once we have consensus what to measure and what data to use for comparison, then we can play with different settings to see how "bad" video can be to still produce satisfactory results.

As you said in today's post, some kind of calibration is necessary. Either standards (crossbar) or pole known length with markers (I like that idea - maybe you can sell it to Essx or Gill).

[chasing6]

One easy reference could be the take off coaches' marks. If they are on the runway they could be a very easy 1' reference, and you could use the same frame to measure grip height. This is all assuming that the marks are accurate (I always double check when I'm at a new track) and that they are present (in most elite vids they are not). Just an idea.

[PaulVaulter]

Everyone can't afford dartfish, but a programme which works basically the same, and is FREE is ImageJ.

It does only work on uncompressed video, but allows you to place digitised markers on you video and take angles and measurements in terms of pixels.

Its not the quickest and easiest progremme in the world to use, but is free, and at least could be used as a double check for your methods your developing Kirk

[KirkB]

No time to work on the Bubka baseline vids this weekend, but here's a synopsis of the Main Vault Phases that I've been referring to in recent posts, where I refer to the "bottom half" and the "top half" of the vault ... plus some new embellishments ...

Phase 1 - the run and plant

Phase 2 - the bottom half of the vault

Phase 3 - the top half of the vault

Phase 4 - the clearance phase

Phase 5 - the landing phase

The main purpose of separating your vault parts into these phases is for (a) common nomenclature; and (b) distinct time-division, so that Video Metrics can be performed on your vault vids. We all understand and agree that as soon as you subdivide your vault into time-based phases, with each vid frame going into one and only one phase, there's some vagueness introduced - and possibly some misunderstanding or disagreement - because we all know that the vault consists of many, many OVERLAPPING vault parts. One purpose of this post is to try to clarify or iron out this potential vagueness so that Video Metrics become more of an accurate science (objective measurements) than an art (subjective or "matter-of-opinion").

Phase 1 is everthing up to the point just before takeoff where your thighs are either parallel to each other, or are about to pass each other. (Good idea, DJ!) This includes your pole carry, run, pole drop, and the substantial part of your plant. It does not include your ENTIRE plant, as you stil haven't PUNCHED or PRESSED your top arm(s) up to finalize your plant. Time-wise, the finalization of the plant coincides with the takeoff, so that goes into Phase 2.

Phase 2 is everything from the "thighs together" frame - which we can refer to as "Frame 0" - on thru to the "body passing the chord" position. To me, Phase 2 is the most critical part of the vault, and the one where I focus my advice ... since I have the most to offer within this phase. In the past, I had referred to this phase as NOT including the takeoff ... and just including everything AFTER takeoff thru to the chord. However, I'm proposing altering this, as I think it's extremely important to have an indisputable Frame 0 reference point. From there, you can count frames forwards 1, 2, 3 ... or backwards -1, -2, -3 ... to communicate which precise frame of a vid you're referring to. Putting the takeoff into this phase (instead of Phase 1) also allows us to separate the energy generated from the RUN from the energy generated from the TAKEOFF ... and beyond!

As 3 further arguments for this proposed separation of vault parts into Phase 1 and Phase 2 ...

1. The "thighs together" frame is like a movie "clap-board", which synchronizes the audio and video of the movie. Similarly, we can use it to synchronize the "Frame 0" frame ... to a very high degree of accuracy.

2. There's many vaulters that have a great run, yet a poor takeoff ... and vice versa, there's some vaulters that have (or in my case, past tense ... HAD) a poor run, yet great takeoff ... so when we talk about putting energy into the "vaulter/pole system" or taking energy out of it, we can more easily separate the two.

3. For Video Metrics purposes, Frame 0 needs to be model-agnostic. For Petrovers, the point in time where the takeoff foot leaves the ground is a perfect way to divide the vault parts time-wise, but then how do you compare it to other models, techniques and styles ... like those that load the pole BEFORE takeoff, or those that don't drive the lead knee on takeoff? You could TRY to see when the foot leaves the ground, but the clap-board synchronization of "thighs together" is so much more accurate ... there's never any doubt as to which exact frame that occurs at. If it occurs MID-FRAME, then take the earlier frame as Frame 0. It's that simple.

Phase 2 ends - and Phase 3 begins - in the frame when your elongated body passes the chord of the pole (or just before this point in time if it occurs mid-frame). Although the chord is an invisible line that can be drawn between the top hand and the butt of the pole (and although the butt of the pole is hidden in the box), you can tell quite easily when your body is most elongated and straightest ... under the pole. This is the body position where your DOWNSWING becomes an UPSWING. It's the position where your WHIP is most powerful/fastest. It's the position where your BACKWARDS C becomes a FORWARDS C (if there is such a thing ... you get the idea). It's the position where your DOWNSWING has put as much energy INTO the pole as you possibly can, and now ... now that you're passing the chord ... your new objective is to invert quickly and get ready to extend your body in unison with the pole's uncoiling.

Phase 3 begins with the start of your UPSWING (as described in the previous paragraph), and ends with your release of the pole by your top hand. Since the pole COILS (bends) during Phase 2, and UNCOILS (straightens) in Phase 3, the primary objectives of Phase 3 are to (a) minimize energy leakage ... by getting the most energy you possibly can back OUT of the pole; (b) get upside down (invert); and (c) extend your body, as the potential energy in the pole transforms to kinetic energy ... thrusting you upwards.

About mid-way thru Phase 3, there's the "flat-back" position ... a potentially HUGE leakage sink-hole. This is the position when your back (even if hunched or tucked) is somewhat parallel to the ground. Your hips have moved up to a level that's even with your shoulders. With Petrovers, you actually pass THRU this flat-back position without the slightest hesitation or pause ... in a single frame. For tuck/shooters (aka drive model or drive vaulters), you stay in this position for quite some time ... often several frames.

Sidenote: The very best elite tuck/shooters have reduced this number of frames down to the absolute minimum possible with their technique, whereas beginners and intermediates ... whether Petrovers or tuck/shooters ... almost ALWAYS stay in the flat-back position for several frames ... almost always far too long.

Phase 4 begins in the frame where you've just released the pole (or if you release it mid-frame, then in the frame just before the time you release it). Other than manipulating your body over the bar, you're at the mercy of gravity for all of Phase 4. It ends when your Center of Mass (CoM) is at it's peak. This peak is not always the same as when you clear the bar. You may clear the bar either on the way up to your max CoM or on your way down (but ideally, EXACTLY on your max CoM), so don't be misguided by the position of the bar when trying to identifying the end of Phase 4.

Phase 5 is your trip down, thru to hitting the mats on your back (hopefully ). There's nothing you need to worry about in Phase 5, other than perhaps ensuring that you don't knock the bar off on your way down, but from a Video Metrics perspective, I think Phase 5 might be important. If you can estimate your max CoM (by knowing the crossbar height), and if you can estimate the height of the surface of the pit above the runway, then you might be able to estimate the number of milliseconds per frame ... based on knowing that the acceleration of gravity is 9.8 m/s^2. i.e. 9.8 meters per second squared.

Once you compute the vertical distance and duration of your landing. then you should be able to compute your camera's shutter speed. Then once you have the shutter speed, you just need to multipy by the number of frames in each phase to get the phrase durations! I'll need Pogo's help with this.

To summarize ... and to tie this altogether with my Bryde Bend vault parts ... we now have:

Run. Plant. Jump. Downswing. Upswing. Fly. Clear. Land.

So now we have ...

Phase 1 - Run. Plant.
Phase 2 - Jump. Downswing.
Phase 3 - Upswing.
Phase 4 - Fly. Clear.
Phase 5 - Land.

I've merely replaced "Whip." with "Downswing." and "Upswing.", and I've added "Land.".

I'm not going to change my tagline, as I still like the sound, energy, and brevity of the word "Whip." ... which includes (as it always did) both the "Downswing." and "Upswing." ... but for Video Metrics purposes, it's necessary to separate the Whip into it's 2 component parts.

This entire Video Metrics idea is about analyzing PV vids by time, distance, and energy, with a particular focus on TIME. If you can determine the TIME each phase takes, calibrate and count the video FRAMES, and you know the vertical DISTANCE to the crossbar (or your max CoM), then ENERGY of each phase can be computed (more or less).

Here's what might become interesting about the energy input and output during these phases ...

PHASE_1_ENERGY_IN + PHASE_2_ENERGY_IN - LEAKAGE = PHASE_3_ENERGY_OUT + PHASE_4_ENERGY_OUT (or something like that)

So if LEAKAGE is a certain constant (constant within vaulter, but variable per vaulter and per vault within vaulter ... which is yet to be determined ... even by Linthorne or McGinness), you should be able to compute your max CoM. i.e. Just looking at your DURATION for vault Phases 1 + 2, PLUS your "somewhat constant" LEAKAGE should allow you to extrapolate your pushoff, thus your max CoM and crossbar height.

Notice that this is without even LOOKING at Phase 3 ... counting frames and computing the duration of Phase 3 is only to confirm your projected pushoff calculation. Just as Linthorne has asserted that your max PR is pre-determined by your Phase 1 speed PLUS exactly 1.00m (his simplification ... to reduce the complexity of his formulae), I'm proposing that you can become even more accurate than that, and pre-determine your max clearance based on only Phases 1 and 2. There's still some things during Phase 3 that can make or break the vault, but assuming those to be fairly constant (i.e. from one of your vaults to the next in Phase 3 technique), then the emphasis now becomes how to improve your Phase 1 AND Phase 2 energy inputs.

You see how I'm stressing Phase 2?

Kirk

[TeamTermin]

Great discussion on measuring.

I teach a Dartfish Sports Movement Class at a community college in western NY. Last semester, as a class project, we did extensive measurement on pole vaulting of elite performers using video downloaded from YouTube, and re-encoded to 30 fps in windows video format.

In your discussions, one of the variables we measured that I didn’t see mentioned was the stride frequency while going down the runway. We measured this parameter on one particular jumper, and found a pretty wide variance between jumps. We measured the stride frequency every three strides from the beginning of the run to pole plant. In addition, we measured the last three strides prior to pole plant, to see if the jumper might be decelerating to set up for the pole plant. In both instances there was a fair degree of variability.

One of you mentioned putting marks on the side of the track, and that would definitely show the variability in velocity. I think if you knew how far back from take-off mark a jumper started, you could measure the stride frequency and the distance per stride, and with the total run time from start to pole plant, back calculate the incremental changes in velocity. For example, this type of analysis is commonly used in sports like swimming for analysis of velocity as a swimmer moves up the pool.

In any case, if any of you have some good video, and have some specific measurements that you want to try to find, we might be able to help. Feel free to contact me . . .
terminb @ ecc.edu