[R] Video Analysis?

[Romain Francois]

On 09/06/2009 01:51 PM, spencerg wrote:
>
> Thanks to Kevin Middleton, Barry Bowlingson, M. Austenfeld:
>
> The Hedrick reference is very interesting, digitizing a video of the
> flight of a moth. This shows the state of the art AND describes "freely
> available MATLAB" code. I've used Matlab before, but it would strain my
> budget right now to obtain it. I've translated Matlab into R before, and
> I could do that again if needed. First, however, I think I want to
> review the "Motion" and "OpenCV" projects -- and possibly "ImageJ".

The RImageJ R package provides some link between R and ImageJ, it is 
very simple (the package has 5 lines of R code), but might help you.

See http://tr.im/qzSl for the annoucement on my blog.

> I want to convert videos into functional time series, more complicated
> than the "gait" data set in the "fda" package but perhaps not as
> demanding as Hendrick's the flight of a moth. I want to measure dynamic
> bridge deflection midspan to see how that varies between spans and
> throughout a duty cycle. I will ultimately want to measure millimeters
> from high resolution video if possible. For a railroad bridge, a duty
> cycle will be the passing of a train, measuring both maximum deflection
> and oscillations. For highway bridges, I want to be able to measure the
> changes from the morning rush hour to midday to the late afternoon rush
> hour -- and preferably also the mostly unloaded period between 2 and 3
> AM. I might be able to place corner reflectors on the bridge to mark the
> points I want to trace, though I'd prefer to be able to do it without
> that. In any event, I'd like to create a functional time series. I may
> also be able to get traffic counts, so I could model deflection as a
> function of traffic.
>
> Thanks again to Middleton, Bowlingson, and Austenfeld.
>
> Best Wishes,
> Spencer
>
>
> Barry Rowlingson wrote:
>> On Sun, Sep 6, 2009 at 1:07 AM, spencerg<spencer.graves at prodsyse.com>
>> wrote:
>>> What software exists for digitizing video to quantify the motion of
>>> specific features in the image? I might be willing to use something
>>> that's
>>> NOT in R, though I'd prefer something in R (or at least with an R
>>> intereface).
>>
>> The 'Motion' project seems to do quite a bit:
>> http://www.lavrsen.dk/foswiki/bin/view/Motion/WebHome
>>
>> Specifically it will capture frames from a video device, detect
>> movement, and write jpeg or mpeg files when it detects movement.
>>
>> You can control it via http commands (which you can send from R) or
>> it can start external commands on events (such as starting to detect
>> motion) - these commands could trigger things in R.
>>
>> Motion is GPL licensed which means you can take the code and do stuff
>> with it - it might be possible to link R directly to the underlying
>> motion detection algorithms.
>>
>> So what's left for R to do? Or do you want to use it to develop new
>> motion detection algorithms?
>>
>> Barry
>>
>
>


-- 
Romain Francois
Professional R Enthusiast
+33(0) 6 28 91 30 30
http://romainfrancois.blog.free.fr
|- http://tr.im/xMdt : update on the ant package
|- http://tr.im/xHLs : R capable version of ant
`- http://tr.im/xHiZ : Tip: get java home from R with rJava