• State TI Thinks Resolved
  • Locked Locked
  • Replies 4 replies
  • Answers 2 answers
  • Subscribers 27 subscribers
  • Views 792 views
  • Users 0 members are here
Support feedback
Options
Options
Similar topics

This thread has been locked.

If you have a related question, please click the "Ask a related question" button in the top right corner. The newly created question will be automatically linked to this question.

DLP4500: DLP4500 pattern frequency selection and embedded vision solution progress

user4686533
Prodigy 155 points
Part Number: DLP4500

Hi Ti experts

We're planning to set up a demo for 3D vision based on DLP solution and I have 3 quesitions as follows:

The first question is regarding how to select DLP type based on our application

  1.  Two 1.3M cameras for stereo vision 3D reconstruction (pixel size 5.3um)

  2. Active structure light will be used, hence DLP solution matches our application

  3. Measurement distance (object to projector/camera) is 1meter with field view about 800mm x 600mm

       P4500 seems to be the right choice, any comments?

       BTW, how to understand the camera resolution shall be 2~3 times of project to achieve good result in TI's training video?  For example, we have 1.3M camera, does this mean 0.4M DMD chip will be sufficient?

The second question is regarding the fringe pitch selection

      
       Three step phase-shifting is planned, any quick guideline regarding the fringe pitch selection? alternatively, how to select the spatial frequency for the projected sinsoidal pattern?

The third quesiton is regarding the progress on TI design tiduc48b (AM5278 EVM+DLP4500) for embedded vision system?

       Phase 2 (part DSP acceleratio)n has already been evaluated, how about phase 3 with Full DSP-acceleration?
     

  • 0
    TI__Mastermind 20045 points
    Hello user,

    1. For our SDK this has been found to produce the best results. For example, if you are using the .45WXGA chipset, the resolution of the DMD is 912x1140 ~= 1MP. Therefore, you would want to find a 2-3MP camera. However, keep in mind that if the field of view of the camera and projector don't exactly overlap (which is normal), you may need to adjust your calculations. The oversampling of the camera is due to the Nyquist theorem. In practice, with different algorithms and different setups you may need to follow this theorem more or less. It is best to determine this parameter experimentally.

    2. Could you clarify what you mean by fringe pitch selection? Are you asking about the pixel pitch of the DMD? As far as spatial frequency of the sinusoidal pattern I would recommend determining what's best experimentally. There are pros and cons of various techniques but this is usually left to the algorithm implementer to determine.

    3. I saw you posted on the other form asking about this question. They will best be able to assist you there.

    Thanks,
    Kyle
  • 0 in reply to Kyle Rakos
    TI__Mastermind 20045 points
    Hi user,

    It looks like my above post answered your question. If not feel free to make another post or ask a follow-up here.

    Thanks,
    Kyle
  • 0 in reply to Kyle Rakos
    Prodigy 155 points
    Thanks for your answer and sorry for the late response.
    We've already ordered DLP4500 and will do some experiments to find it out.
    As for fringe pitch selection, I mean how many DMD pixels for each sinusoidal period. I find this is really a trade off between resolution and phase unwrapping complexity. We'll try to do some experiment to find it out too.
  • 0 in reply to user4686533
    TI__Mastermind 20045 points
    Hi user,

    Thanks for the reply. Best of luck!

    -Kyle