Markless (Manual) Digitizing

Hello Tamra,

 

Thank you for your message.  Yes, in theory, you are correct.... if it takes one minute to digitize a single image and you wish to digitize one second of video at 60 images per second, then it would take 60 minutes to digitize this sequence.  However, while this works out mathematically, this is not a realistic scenario.  I will explain below.

 

The "manual" mode of digitizing is also referred to as "semi-automatic" digitizing.  In the first image of a sequence, the software does not have any idea where the desired points are located, therefore, the user must move the cursor to each point and press the mouse button to digitize that point.  The APAS software incorporates a prediction algorithm that will position the cursor in the expected location based on previous information.  In the second image, the software only has one previous image to work with (Image #1) so the software will automatically place the cursor in the same location for each point that was digitized in the first image.  The user then makes any minor adjustments (if necessary) and clicks the mouse button to digitize the location of the point in the second image.  When the image is advanced to the third image in the file, the software now has two previous digitized locations to work with (Image #1 and #2) so the software uses both position and velocity measures to predict the location of each point in the third image.  As with the second image, the user makes any minor adjustments and clicks the mouse button to confirm the location of each point.  After 3 or 4 images have been digitized, the prediction algorithm uses position, velocity and acceleration measures and works very well and the user is only clicking the mouse for the digitizing process.  This "semi-automatic" mode greatly increases the speed of the digitizing process.

 

What this means for practical purposes....  An experienced user can manually digitize a full golf swing, or tennis swing from 3 cameras (60 images each) in 15 or 20 minutes.

 

There are other options that can be used to increase the speed of digitizing, though the use of these will depend upon your exact applications.

 

    -  The APAS does not require that every image be digitized.  The user can digitize every second or third image during portions of the video where the movements are relatively slow.  For example, on a simple gait analysis, if the subject is walking very slow, then the user could digitize one image, skip one or more images, digitize another image, skip one or more images etc....  The APAS will automatically use a linear interpolation to fill in the missing data

 

    - The APAS-XP software option includes the CAPDV software module where video can be captured directly to the computer hard disk drive from up to five cameras simultaneously.  This significantly reduces the time required to capture video individually from multiple cameras and then synchronize the video together.

 

I hope this helps to answer your questions.  Please feel free to contact us for any additional information.

 

Sincerely,

 

John Probe

Ariel Dynamics, Inc.

Email:  ARIEL1@ix.netcom.com

 

---

Hello Yang Bin,

Your problem is a very difficult one. We do have some success with automatic digitising using the APAS system. As far as I am aware APAS is the only system that uses black markers so you will have more success wioth this system than any other. Another advantage is that even when you lose the marker for a few frames you can digitise manually until it starts picking them up again. In addition to the problem with bubbles we have found that the blue filtering effect of the water reduces contrast. There are several things you could try:

1. Try changing the size of the black markers.
2. Increase the contrast by using a white background against the black marker.
3. Increase the sensitivity of the software to detecting diferences between black and white
4. Have the cameras closer to the swimmer.
5. If you are using 3D methods have the camera deeper. The markers may be less obscured by bubbles when looking upwards rather than from the side.
6. Try using black balls attached to the swimmer. Black balls from hobby shops can be useful - We attach them using a hot melt glue gun to several layers of white surgical tape.  

We have tried uisng MATLAB to identify pixels of contrasting colours (differences in RGB content) to overcome the effect of the blue filtering that reduces the contrast of black and white. However, we did not get as far as finding a system that worked.  

Please do not hesitate to contact me if you would like to discuss the problems further.

Do you know Shuping Li? He has just finsihed his PhD at Edinburgh and is now back in China.

Good luck with the swimming analysis.

Ross


—Ó ±Û <yb1017@hotmail.com> writes:
Dear Mr. Sanders:
    How are you?
    I am Yang Bin, an technician of Beijing F&H Co, China. We met at the
International Sports Biomechamics Conference in Beijing last year .
    Now we are dealing with the matter of National Swimming Team of China.
We use the cameras under the water to capture the video data. They want to
use the automatic digitize. We paint the angle points with the black marker
pen. But when the athlete's speed is a little high, there are too many air
bubbles. In the water, the air bubles are white. We cannot see the points
clearly, so we cannot digitize automatically.
    How can we resolve this problem?
    Thank you very much!

                                                         F&H, Yang Bin

_________________________________________________________________
ÏíÓÃÊÀ½çÉÏ×î´óµÄµç×ÓÓʼþϵͳ¡ª MSN Hotmail¡£  http://www.hotmail.com  

_____________________________________________________________________________________________________

Hi Guys,

Bob - that was it (turning off the missing pt box). Went a lot faster. I digitized your cube with 8 points in the order we do for our baseball batters.

Had a few bad points toward the end, but passed along the .3d file to Rich (16 pts) to play with on the skin. We'll need to make it work for different xyz coordinate spaces for sure - probably just a toggle we'll put in.

Thanks guys,

Scott

---- Original Message ----
To: <sbendis@virsys.net>
From: "Robert Wainwright, P.T." <kincon@bellatlantic.net>
Date: Thu, 23 Dec 2004 09:59:45 -0500
Subject: Re: Auto Digitizing Question

Hi Scott
Did you digitize the calibration cube or did you copy (either is okay but must be done before using digi xp)? Did you get all 5 video's to appear on the screen automatically after selecting the proper first file? Second, you must turn off the notify if missing point box under auto digi options. It is using the new algorithm and only notifies you if unable to see two, otherwise, it askes for help on all missing points, all pictures. I suggest you open my completed file, note the settings and use for yours. There will be some frames that need help if not seeing two points or extreme acceleration but it should be good for 95% or more. It is all in the settings.

Let me know how it goes.
Bob

----- Original Message -----
From: <sbendis@virsys.net>
To: "Robert Wainwright, P.T." <kincon@bellatlantic.net>
Cc: "GideonAriel" <gideon@arielnet.com>
Sent: Thursday, December 23, 2004 7:12 PM
Subject: Re: Auto Digitizing Question


> Hi Gentlemen,
>
> I'm trying my first attempt at using Auto Digitizing on Bob's Golfer (5
videos). For every frame there are points that it does not see (which I'll
click on manually to help finish) but it does this for every frame.
>
> Is this the process you need to help it go through? or do you hit escape
or some other feature to make it go faster?
>
> Thanks Guys,
>
> Scott

 

---