1 (edited by Chas Tennis 2016-07-13 16:48:39)

One application that I am interested in is comparing tennis strokes on side-by-side screens.  I often have video clips of a tennis server and want to compare them to high level severs side-by-side.   

Sometimes the videos have been downloaded from Youtube, Vimeo or other sources.  Sometimes the recording rates stated by the originator of the video are questionable. So I want to better understand how Kinovea handles timing for side-by-side videos.

I'm using Kinovea version 8.15. If a later version has improvements for coordinating time scales on side-by-side videos, please mention.

In tennis, some researchers use a time scale with ball-racket impact as time zero.  Time or frame count values before impact are negative and those after impact are positive. I want to display positive and negative frames & times in this way but have not found a way to display negative frame rates or times. I have counted back from impact many frames, 130 frames, and placed the zero frame there but would prefer using negative values before impact.

Example of a very informative time scale, displaying joint motions for a tennis serve using impact at time zero.  Impact at zero seems appropriate for all striking sports so negative frames and times would be an informative number for Kinovea to display. 
http://g4.img-dpreview.com/70E9CDAB577B4335859535EAF3EF5940.jpg

Just starting these side-by-side comparisons. I recently analyzed an iPad Air 2 video 'recorded at 120 fps' and another video recorded at 240 fps. When I enter 120 fps for the iPad camera and 240 fps for the other, the ball seems to reach the same positions after impact frame for frame.  That's confusing. ? Was the one serve twice as fast as the other?  I guess that Kinovea simply assumes that all frames originally recorded are present in the video being analyzed. ? 

Can the entered recording frame rate be changed at any time during analyzing the video? Or does it become locked at some point, say, after the first key image has been saved?   

----------------------------------------------------------------
Information in Kinovea 8.15 "Help".

http://www.kinovea.org/help/en/115.html

"Configuring for high-speed cameras

High-speed capture - Slow motion playback

High-speed camera record videos at a very high frame rate and store them in a video file with a normal frame rate.
For example, a high-speed camera may record images at 1000 frames per seconds and create a video specifying that it should be displayed at 30 frames per seconds.
When played back, the video will appear in slow motion.

By default, the times displayed in Kinovea are based on the playback frame rate, since it is the only information encoded in the video.
Hence, when playing back a video captured with a high-speed camera, the times displayed will not match the captured event time.

You can specify the capture frame rate of the high-speed camera by right clicking the image to bring the main contextual menu and then use Set Original Speed…


Configuring the high-speed camera recording frame rate.

All time-related informations displayed in Kinovea will use this information (i.e: current position, working zone duration, stopwatches, etc.)."

Motion - high speed video -
http://www.kinovea.org/help/en/115-highspeedcameras.png

---------------------------------------------------

I've noticed that the frame counts can be off between the two windows if I start and stop the video rapidly.  I assume that is a computer processing time issue. ?  If I slowly press the pause control to advance single frame, the frame counts compare in a more well behaved way.  Suggestions to users when frame rates do not agree when advanced rapidly?


[Someone mentioned a way to roughly measure video frame rates.  Was the frame rate 25 fps, 30 fps, 200 fps, 240 fps, etc.?  I believe that if a ball is bounced and the peak of the bounce is videoed for a particular type ball, it will always drop a certain number of diameters in a given time.  For example, a  tennis ball or soccer ball, etc. could be used as a time reference and may even appear incidentally in a video. This time reference might be useful for double checking video frame rates.  For example, videos could be made showing the peak of a tennis ball bounce for different video frame rates.

Some smartphones are said to change frame rates in low light, to use slower frame rates to collect more light.]

2 (edited by joan 2016-07-13 19:00:06)

This is definitely handled differently in 0.8.24.

When using comparison the videos should be both progressing relatively to a common absolute time. There is also an option in the preferences to unlock the speed sliders if that's required. The speed slider percentage is relative to real time (takes capture framerate into account if it has been set).

After the synchronization point has been set between videos, the times before it are indeed expressed as negative.

0.8.25 also has a additional setting to force a different reference playback framerate, in case the video metadata is wrong.

It's also my experience that most "cheaper" sensors will auto-exposure in low light conditions and degrade the framerate to a value of 1/exposure duration.