Jump to content
Sign in to follow this  
SweatyYeti

Patty's Arms And Hands

Recommended Posts

MikeZimmer

Thanks,

 

I will look at this in the future. Now, bedtime, and tomorrow - get ready for X-mas.

 

Regards

 

 

I had a quick look. I am not sure yet why it works, but will play with it, and apply it to arm swings - should be able to. The big difference is that the figure itself contains the centre of rotation - should not be an issue of course.

 

Time to dig out ruler and compass.

 

I want to do some experiments with clear images of people, to see how various techniques might work.

 

One thing that strikes me is that given how massive Patty's upper arms seem, it might add error to where to place the humerus between triceps and biceps. More area - more room for error?

Share this post


Link to post
Share on other sites
MikeZimmer

I had a quick look. I am not sure yet why it works, but will play with it, and apply it to arm swings - should be able to. The big difference is that the figure itself contains the centre of rotation - should not be an issue of course.

 

I tried some experimentation. The method seems to work quite well for determining centre of rotation of a rotated figure, when the centre has not moved. The centre of rotation can certainly be within the object. You only need two points to be established on the object, and these must be established identically on both images, in the same place exactly to minimize error. They should cross the radius defined by the centre of rotation, and not line up with it. If we could line up along a radius exactly, we could use another method. However, this almost begs the question of where the centre is.

 

I don't know how to do this electronically, since I don't have any idea if any tools I have would simulate drawing an arc with a compass. This seem to force work onto a printed copy.

 

I used Libre Office just now. I try not to buy software if I can avoid it, and use free and open source where feasible. I don't see an option to upload an image that is not at a URL, so my diagram is not here.

 

 

Determining Centre of Rotation with Images

1 – assume that we are dealing with a two dimensional object

2 – the image is not foreshortened

3 – there is no translation (i.e.) no straight line motion, so the centre of rotation does not move

4 - There has been sufficient rotation - with too little rotation, it will be hard to get accurate results I think

Using a geometric solution, with ruler and compass

 

Mark two points on the first image, on opposite sides, and call them A and B

 

Mark two points on the second image, in the identical positions as on the first image, and call them A' and B'

 

With a straight edge, draw a line between A and A'

 

With a straight edge, draw a line between B and B'

 

Set the point of a compass at point A and make arcs above and below the A to A' line, then repeat for point A'. Make the arcs long enough that they intersect and are above and below the line where they intersect by a distance roughly equal to half the length of the line. Join the points where the arcs intersect on one side of A to A', to where the arcs intersect on the other side of A to A' with a straight line, using a straight edge. This gives the perpendicular bi-sector of the line A to A'. The centre of rotation will lie somewhere along this line.

 

Follow the same procedure using the B to B' line, to make the perpendicular bi-sector of the line B to B'. The centre of rotation will lie where this line intersects the perpendicular bi-sector of A to A'.

 

You don't actually have to draw the lines A to A' and B to B' to find the centre of rotation. This was stated to help you visualize the perpendicular bi-sectors.

  • Upvote 1

Share this post


Link to post
Share on other sites
SweatyYeti

^

 

That's an interesting exercise, Mike. I may give it a try myself, at some point. :)

 

Right now, I'm working on another animation...for comparing Jim McClarin's and Patty's walks.

Share this post


Link to post
Share on other sites
MikeZimmer

^

 

That's an interesting exercise, Mike. I may give it a try myself, at some point. :)

 

Right now, I'm working on another animation...for comparing Jim McClarin's and Patty's walks.

 

I am trying to find good images to use, but it is harder to find clear videos taken from the side with arm swing than I thought. Also, my video down loader - worked flawlessly for years - now returns empty files. It may be my anti-ad software. 

 

I have made some progress in understanding the technique, associated limitations, and how to improve the method. It turns out that what you are doing is creating a chord for a circle, and a chord for another concentric (same centre) circle, creating the chord bi-sectors, and seeing where they meet. You can add more reference points - they all should converge to one centre point. They won't if there is translation of the figure and not just rotation. If you have forgotten your geometry, a chord on a circle is a line joining any two points.

 

You may find that your drafting skills result in some inaccuracy as well - mine do.

 

The technique is really much like looking at a time exposure of a night time sky, to the north, with circular star trails - you probably have seen pictures of these.

Share this post


Link to post
Share on other sites
xspider1

I'm glad you ran with the method above, Mike.  Don't know if you have this series of images but, perhaps there are good candidates for determining the centers of rotation (shoulder joint and elbow joint) here: 

 

post-131-0-10157500-1419610895.gif

 

I captured 4 frames from that gif:

 

post-131-0-32031300-1419611515_thumb.jpg

(click to enlarge)

 

I think one can upload images here after 50 posts (not sure).  I'd be very interested in seeing your results. 8 )

Share this post


Link to post
Share on other sites
MikeZimmer

Thanks,

 

I wrote a different post, but hit the wrong key and lost it apparently.  Short version, it was already done this way with rotation centres in Legends Meet Science, with professional video software, and the intermembral index was in the range 85-90. I don't really know what that means in terms of confidence, but let's assume it is at least a 90% confidence interval. The human intermembral index is around 71, with a smallish standard deviation - tight group, around 2 intermembral index units or so. The chances of the creature being human are miniscule, based on these figures along. This agrees with some of Sweaty's calculations.

 

 

I found that I actually own the book by Green discussing this (briefly, in sketchy fashion). Must have read it with little focus.

 

 

I have started to write a summary paper discussing the Legends Meets Science work on intermembral index from a statistical, controlled research perspective. I covered the basics of my approach in three articles posted previously, but continue to get a clearer handle on what the issues will be from the viewpoint of scientific adequacy.

 

I would like to talk to John Green on  this, but understand that his health is not good. He lives pretty close to me, but I hesitate to bother him. I have not looked to see if he has a listed land line number.

 

I don't try to address the fact free assertions coming from certain quarters, both on and off this list, but do try to assess where legitimate objections may arise, and what is required to address these.

 

So stilts, arms extensions, scaffolding, ..., fact free assertions

 

Error of measurement, variability, confidence intervals, experimental bias, foreshortening, image quality, ..., reasonable objections to be met.

 

 

The hard part for me is the actual measurement - I have no skills with graphics software.

Share this post


Link to post
Share on other sites
MikeZimmer

I'm glad you ran with the method above, Mike.  Don't know if you have this series of images but, perhaps there are good candidates for determining the centers of rotation (shoulder joint and elbow joint) here: 

 

 

 

Still working on how to do it. my current thoughts are below.

 

 

Geometric Location of Centre of Arm Rotation at Shoulder

Preliminary Notes on Method

 

Introduction

 

In order to measure the length of a bone in vivo, or in images, it is necessary to know where the ends of the bone lie within the surrounding soft tissue. One method that has been used successfully in the past is to identify the centre of rotation of the two end joints of the bone. This has been done using advanced video processing software, run by skilled experts. This paper explores a simple geometrical method based on the analysis of video frames, to find the centre of rotation of the humerus, at the shoulder joint.

 

General Approach

 

1 – Obtain appropriate video stills showing rotation of the arm about the shoulder socket

2 – Registration of images to insure that they share the common centre of rotation

3 – Marking of common points for analysis using geometric techniques

4 – Calculation of centre of rotation using ruler and compass

5 – Analysis of errors resulting from procedures and setting some probabilistic bounds

 

 

Details of Approach

 

Obtain Images

 

Stills obtained from clear video of runners will be used as a test.

There will be one still with the right arm to the rear, and one still with the right arm to the front

 

Register Images

 

Here we are presented with two related images, coming from a video where the arms are pumping, the legs are scissoring, the torso twisting, and the shoulder girdle moving. There is a centre of rotation at the shoulder joint, a ball and socket joint. The precise location is obscured by the soft tissue, and since the whole body is moving, running, the centre or rotation is moving as well. In order to make a geometric determination, the two images, arm back, and arm forward, must be registered, overlaid, so that they have one point for the centre of rotation. Otherwise, the geometric solution will produce an erroneous result. Here we end up with a catch 22: in order to determine the centre of rotation, we must know the centre of rotation to register the images. If this were the whole story, we could quit now. However, it is possible to register the images, perhaps with some bounded error, by looking at the part of the shoulder above the arm, and trying to match locations. There is a clear possibility for error of course, and this will be circular error. We can put bounds on this, and test the calculations with the bounds to see the effect on location of the computed centre.

 

Marking of Common Points

 

If three points are marked on the limb in the rear position, the same, congruent, three points must be marked on the limb in the front position. Errors between the two limb positions will create error in the results. The calculations can be done with only two points, but three create an extra check.

 

I suggest that an equilateral triangle be used, with one vertex near the point of the elbow, one perpendicularly opposite that on other side of the upper arm, and the remaining determined by the three 60 degree angles. Other configurations may be tried.

 

Calculation of Centre of Rotation

 

The calculations are actually done with ruler and compass on printed matter. The vertices on the figures are lettered A, B, C on the rear arm position, and A', B', C' on the front arm position. Lines are drawn between corresponding letter pairs: A to A'; B to B'; C to C' . Arcs are drawn from each vertex and the perpendicular bisectors are drawn for each of the three lines. Where these bisectors cross will be the centre of rotation if your registration was accurate, and you geometric drafting well done. If they do not cross in one point, they should all come near to each other. You can check your work to see if you can improve the result by improving your drafting. If that is not the issue, it means that the registration is a bit off. You can play with this, but I can not be more specific at my current stage of knowledge.

 

Analysis of Errors

 

 

There are a number of potential sources for error in the the process. you may be able to minimize these, and set some bounds, probabilistic bounds perhaps, on the effect of the rest. These may be classified as follows:

 

1 – image quality

 

2 – registration errors

 

3 – marking errors

 

4 – drafting errors

 

5 - Foreshortening

 

Image quality will affect the accuracy of results. Steps 2 and 3 both depend on selecting the correct spots to mark on the images. If the boundaries are fuzzy, it becomes harder to get the correct, congruent spots. The registration process is going to be subject to error. It should be possible to set some bounds on the size of this error.

 

The geometrical figure drawing is very straight forward, and the accuracy of this step can be improved with practice. It should not be a limiting factor.

 

Foreshortening is the result of projecting an image onto a surface when the angle of the object creating the image is not coincident with the plane of the surface. Maybe someone else can express it more clearly.

 

Foreshortening of an image will definitely change the calculation of the angle of rotation. I have not determined if it will change the centre of rotation determination. If it can, it may not be possible to determine the centre of rotation using this method.

Share this post


Link to post
Share on other sites
Trogluddite

^^

One quick thought - if you could get pictures of known individuals under controlled circumstances, you could do the video analysis then actually measure the arm/leg to determine the margin of error for this methodology.  

 

To control against "knowing" the answer you need to hit, maybe work with a photographer who knows what you need in the images.  That person photographs the subjects and provides detailed information on the camera settings/conditions, but does not provide the information on the subjects who were photographed.  

 

Analysis is done, then the subjects come forward to be measured.  

 

Once done with an athlete were the limbs are visible (e.g., runners in shirts w/o sleeves), test again with individuals whose limbs are less visible - for example, people walking in hooded sweatshirts. 

 

This seems like a great project; I'm sure that the more serious individuals on the forum would not mind assisting. 

Share this post


Link to post
Share on other sites
MikeZimmer

^^

One quick thought - if you could get pictures of known individuals under controlled circumstances, you could do the video analysis then actually measure the arm/leg to determine the margin of error for this methodology.  

 

To control against "knowing" the answer you need to hit, maybe work with a photographer who knows what you need in the images.  That person photographs the subjects and provides detailed information on the camera settings/conditions, but does not provide the information on the subjects who were photographed.  

 

Analysis is done, then the subjects come forward to be measured.  

 

Once done with an athlete were the limbs are visible (e.g., runners in shirts w/o sleeves), test again with individuals whose limbs are less visible - for example, people walking in hooded sweatshirts. 

 

This seems like a great project; I'm sure that the more serious individuals on the forum would not mind assisting. 

 

 

I may end up experimenting with controlled images made by myself, to see if it is really doable without in essence begging the question of where the centre of rotation is. This whole bit about registration has me buffaloed.

Share this post


Link to post
Share on other sites
MikeZimmer

I may end up experimenting with controlled images made by myself, to see if it is really doable without in essence begging the question of where the centre of rotation is. This whole bit about registration has me buffaloed.

 

 

I am done experimenting for now, since I am not convinced that the technique is a workable one. There is too much guess work involved in setting up the point for registration of two images, and also in marking the arm. The arm seems to be wider in one image than in the other, and I guess it is due to the different videoing perspectives taken. In any case, I don't see this as a feasible technique. Perhaps someone with some actual training in measuring photographs could succeed. It was an interesting experiment, and led me to start figuring out how to use the GIMP image manipulation program.

 

Vision Realm Entertainment used some specialized software in their effort, seen in Legend Meets Science video. They did their analysis work around centres of rotation according to John Green. Clearly they have a better way to establish this.

Share this post


Link to post
Share on other sites
Trogluddite

MZ, 

 

Sorry to hear that you hit a wall, but if you're ever in [REDACTED], NY I'll by you a beer for trying.  I won't claim to understand the technical photography stuff as I only had time to do a cursory skim of your posts, but the logic of what you were attempting to do was clear and appeared sound.  Maybe someone can pick up the ball and run from here or collaborate with you to find a resolution to the technical barriers.

 

Again, tho, two thumbs up for putting a shoulder to the wheel. 

Share this post


Link to post
Share on other sites
MikeZimmer

MZ, 

 

Sorry to hear that you hit a wall, but if you're ever in [REDACTED], NY I'll by you a beer for trying.  I won't claim to understand the technical photography stuff as I only had time to do a cursory skim of your posts, but the logic of what you were attempting to do was clear and appeared sound.  Maybe someone can pick up the ball and run from here or collaborate with you to find a resolution to the technical barriers.

 

Again, tho, two thumbs up for putting a shoulder to the wheel. 

 

 

Thanks fro the encouragement. I don't consider it a bad investment of my time - think of the Edison lightbulb research model.

 

Michael

Share this post


Link to post
Share on other sites
MikeZimmer

Brachial Index for Patty Based on Measurement of Vision Realm Skeleton

 

Just a note about work I have been doing on systematically analyzing limb proportions using various data sources.

 

Using the skeletal image of Patty produced for Sasquatch: Legend Meets Science, by Vision Realm Entertainment, I have calculated four indexes and statistically compared them to human benchmarks:

 

  • Brachial

  • Humofemoral

  • Crural

  • Intermebral

 

The relative image measurements were based on bone lengths (in pixels) and the last three ratios were many standard deviations away from the human mean. Not a chance they reflect a human being. The exception was the Brachial Index, which was only slightly more than two standard deviations away from the mean.

 

The Brachial Index value does not line up with the calculations done by SweatyYeti. I get 81.19. Sweaty gets 91.8 I believe. Below I suggest reasons why this may be the case.

 

See post 1169 on this thread (http://bigfootforums.com/index.php/topic/4782-pattys-arms-and-hands/), Patty's Arms and Hands. Excerpt here:

 

“SweatyYeti, you have computed a humorofemoral index of 91.8. The human humofemoral index lies in the range of 69.8 to 72.8 according to one source (referenced in a previous post). The standard deviation for humofemoral index for humans lies in the range of 1.79 to 3.41, depending on which study you read. All of these numbers were based on small sample sizes, which is really too bad. There may be better numbers out there; the trick is to find them.

 

The smallest difference from the list of possibilities is 91.8 - 72.8 = 19.0, based on a sample of ten Caucasian males. The corresponding standard deviation was 1.62. This difference represents 19.0 / 1.62 = 11.7 standard deviations. Using standard probabilities for standard deviations, the likelihood of this being human is vanishingly small.â€

 

My results may reflect:

 

1 – Errors in the original work by Vision Realm

2 – Distortion of the proportions due to various duplication steps on the image

3 – Errors in my measurements

4 – Errors in my calculations

5 – The difference between attempting to measure soft tissue images instead of skeletal images

6 – Errors in SweatyYeti's calculations – done very differently.

7 – Other factors as yet unknown to me

 

I also modified the anthropometric reference data slightly, averaging to include more observations. One thing that I have noticed is that the reference data is not as consistent as one would like. This is in part due to different methods of locating the endpoints for the measure, soft-tissue, bone, different landmarks, and partly that the reference samples are not pooled for mankind in general, but reported for very small groups. Sometimes, the number of samples and the standard deviation are not given. A lot of the material is available in books, but subject to the limitations of Google Books.

 

The bottom line is that there is a lot of inconsistency in the science, and maybe a specialist in the area could give clearer guidelines.

 

The full report is on my thread at http://bigfootforums.com/index.php/topic/49888-biometrics-and-the-ontological-status-of-sasquatch/#entry884721, post 17. I would be pleased if someone could take a look and tell me of deficiencies.

  • Upvote 1

Share this post


Link to post
Share on other sites
Guest

I am done experimenting for now, since I am not convinced that the technique is a workable one. There is too much guess work involved in setting up the point for registration of two images, and also in marking the arm. The arm seems to be wider in one image than in the other, and I guess it is due to the different videoing perspectives taken. In any case, I don't see this as a feasible technique. Perhaps someone with some actual training in measuring photographs could succeed. It was an interesting experiment, and led me to start figuring out how to use the GIMP image manipulation program.

 

Vision Realm Entertainment used some specialized software in their effort, seen in Legend Meets Science video. They did their analysis work around centres of rotation according to John Green. Clearly they have a better way to establish this.

 

Hi Mike,

 

You must derive the articulation points. The easiest way to do this is to go old school and use an adjustable mannequin and develop a model by matching Patty's orientation and distances from the camera for frames 288 to 480. Adjust the dimensions as required to fit the model. This frame sequence represents 16.8 steps of the clearest, most consistent film footage and Patty is essentially a walking ruler @ approx. 12 frames per step.

 

Else if you have the software, like VisionRealm, you can create a framework model which overlays the subject EXACTLY for every undistorted frame of the PGF. The data will consist of geo-coordinates of the pixel vertices for the vector framework which allows you to derive the 3D vector lengths and their corresponding foreshortening. This is the essential data required to model Patty over the trackway which is similar to tracking the white balls they put on actors to render them in CGI.

 

While creating the model you fine tune the framework dimensions until they converge on a best fit of Patty's actual body dimensions. This also constrains the error. After you converge on a solution you can easily measure the intermembral indices from the model to a known degree of accuracy. However, we first must establish the framework's vectors and vertices which include the articulation points on the body. Here is an example framework:

 

post-337-0-26655700-1420583158.png

 

This framework is stored as 3D joined vertices known as a complex chain. Each vertex holds the x/y/z location including its position in the chain. Then we can derive the foreshortening of each vector and calculate the actual length of each body part. In all, it's a lot of work for likely little reward. Skeptics won't accept it if:

 

1) They don't understand it.

2) They don't like it.

 

Which covers most of the bases. I hope to have more time in the new year to work on this. Until then.

 

Happy New Year!

GF

Share this post


Link to post
Share on other sites
xspider1

^ ^^ Great stuff.  I really like the idea of using an adjustable mannequin.  That should definitely work!

 

:good:

Share this post


Link to post
Share on other sites
Sign in to follow this  

×
×
  • Create New...