Show movement in area from 30m video, smart objects - stack mode - range - 90k frames

Question

Hey,
So I have a real challenge and need help figuring out the logistics of how to get it to work sensibly.
I'm trying to visually show the movement through a space (which will later be generated into a 3d form etc..)
To do this I think the best way would be to use photoshops smart objects - Stack modes - "Range"
A final image to show movement through the area.
Lighter where people walk more often and darker where people walk less often
http://help.adobe.com/en_US/Photosho...44D6937Ca.html
Range basically shows everything that has moved between the layers.
If you import a video frames into layers in photoshop, make them a smart object and apply the range stack mode it shows peoples movement and where they walk.
So thats what I'm going to do unless there is a better way to do this in something like after effects or premiere pro that I don't know about?
- I've done the photoshop way before and it worked very well - but on a very small scale.
I have a 30 minute video recording (considering longer but not sure if its viable), it was shot on a tripod and seems ok but can always do it again differently if needed!
The video was shot at 1920 x 1080 50fps which with 30 minutes (1800 seconds) makes 90,000 frames total for the video.
The problem
I need the info from ALL the frames - any less and the movement is too 'choppy' and isn't very useful.
So how can I make use of 90,000 frames/layers in a smart object for the range stack mode?!
I heard photoshop has a layer limit of 8k.
I have access to 6 high spec brand new mac pros (not 100% sure of the spec).
I'm thinking the solution would be use scripts and break it down into manageable chunks then flatten then merge them together.
Possible Solutions
1. Merging frames from each second or few seconds and then blending - not sure what blend mode I would use and also not sure if it would affect results
2. chop the image down into sections, process each section, flatten and merge together. for example split it into 4 quarters. - Don't think this will help, its more the limits in layers than anything else
3. Find an alternative method to deliver the image such as after effects.
on a side note I know I could have done a long exposure, done many before, but they don't really show movement, they simply blend the frames. The downside of long exposures for this exercise is that things could cancel each other out, for example; a man with a white top on walks past then a man with a black top walks past, these would cancel each other out to some degree.
The Range smart stack mode is brilliant because it directly shows movement and people don't cancel others out.
Help
Where can you help?
So basically if this is the correct way to do this then can anyone help with this and suggest any ways of breaking it down that wouldn't affect the final result?
Someone that knows more of how range stack mode works.
Correct me if I'm wrong with any of this and please let me know if you can help at all
Thanks!
Luke.
Context
For context the final thing has to be something that doesn't change in time, i.e. not a video but an image or model. This image will then be skewed a bit so its a plan view. It will then be used for another script in a 3d package like rhino which will create a number of points in space, the x and y axis being the location in plan view and the Z axis being the value of the pixel. I will then generate a solid form from the points and then smooth it out a lot so I should end up with a 3d wave model directly generated from the movement of people through the site. I will then architecturally design the form into spaces etc... This will form the surface of my final project
Heres an example of the first 500 frames (500 frames limit on video frames to layers import)
The footage isnt the best as I moved the camera slightly and the angle isnt amazing - ideally I want a plan view but will skew the final image to do this.

Answer 1

But I have a concern that if I break it down to 300 images (6 seconds of footage) then stack then flatten the stack until its dont the whole 30 mins then use these flattened ones to create a new stack it would affect results as it will then be working with images like the above.
That should be easy enough to test (if maybe not exactly proof) – for Range it seems to make no difference, for Variance it seems to do.
Also why use 16 bit image mode?
To reduce the accumulated effect of the necessary rounding of pixel values.