From: michael.balda@gmail.com   
      
   Am Samstag, 1. Februar 2014 09:50:46 UTC+1 schrieb haitic...@gmail.com:   
   > On Friday, January 31, 2014 6:18:23 AM UTC-5, Michael Balda wrote:   
   >    
   > > I'd like to use a pattern projector for computer vision tasks. The pattern   
   can be in the NIR range, should not repeat itself but doesn't need to be   
   described exactly, so random noise, speckle etc. should be fine. It would be   
   preferrable, if the    
   pattern can be restricted to a defined field of view instead of being   
   omni-directional and was stable over time. Of course there are several   
   different approaches but I would like to find one which is    
   >    
   > >    
   >    
   > > (a) cheap,    
   >    
   > >    
   >    
   > > (b) robust,    
   >    
   > >    
   >    
   > > (c) efficient (in terms of power consumption),    
   >    
   > >    
   >    
   > > (d) not too complex,   
   >    
   > >    
   >    
   > > (e) has good availability of components and    
   >    
   > >    
   >    
   > > (f) is monochromatic or has small bandwidth.   
   >    
   > >    
   >    
   > >    
   >    
   > >    
   >    
   > > Thanks in advance for any suggestions.   
   >    
   >    
   >    
   > There's random and then there's random. Ditto for monochromatic. It would   
   suit the discussion if you could say more about what you are using it for. I   
   like the multiple led or laser diode solution best because it has no moving   
   parts. In this approach,    
   the Picaxe, a $5 chip, outputs PWM (Pulse Width Modulation) pulse train to say   
   5 led's separately. This runs in the background on the Pic. Very easy to do.   
   The randomness is combinatorial in nature, so even if you have each led glued   
   to a ground glass    
   scattering screen, and even if each pattern is fixed, the combinations are   
   very random, as the random number generator on the chip sets each led   
   differently.    
   >    
   > I believe there is a "rand" function on board, but I am in habit of running   
   a few times to prime pump, then filling an array with the "random" numbers,   
   and then using subsequent random numbers to index into the array.    
   >    
   >    
   >    
   > Perhaps this is a "thought experiment" on your part. Anyway, fun to think   
   about.   
   >    
   > jb   
      
   Thanks for the suggestions. The purpose is to generate an intensity   
   inhomogeneity in camera images even on homogeneous surfaces. This can be very   
   useful for finding  correspondences in stereo-vision. Repetitions can yield   
   false correspondences, so that's    
   why the pattern should be pseudo-random. Since high contrast even in daylight   
   would be nice, small bandwith and efficiency are an issue.   
      
   --- SoupGate-Win32 v1.05   
    * Origin: you cannot sedate... all the things you hate (1:229/2)