• State TI Thinks Resolved
  • Locked Locked
  • Replies 9 replies
  • Answers 1 answer
  • Subscribers 30 subscribers
  • Views 951 views
  • Users 0 members are here
Support feedback
Options
Options
Related

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.

DLPLCR4500EVM: Poor results with acquired Point Cloud

John Traynor
Intellectual 885 points
Part Number: DLPLCR4500EVM


Hi,

I have the following setup 

EVM4500

Microsoft Lifecam Camera

https://e2e.ti.com/cfs-file/__key/communityserver-discussions-components-files/94/LCr4500_5F00_3D_5F00_Scanner_5F00_Application_5F00_v2.7z

I have calibrated and tried to perform a scan on a simple box. 

The results are very poor so I believe something has gone wrong in calibration but I'm not sure what. Calibration values for both the projector and camera were below 2.

Could someone have a look at the files I have attached and see where I am going wrong. I basically have attached all the calibration images and some output scans

Many thanks

JT

  • 0
    TI__Mastermind 30875 points

    Hi JT,

    Please try the following options -

    1. Perform only vertical scan. See the results

    2. Perform only Horizontal scan. See the results.

    The setup is performing a graycode scan for 640x480 resolution, from your scan results, in the \LCr4500_3D_Scanner_Application_v2\output\scan_images\ directory data, i can see the fine lines both horizontal and vertical lines almost not visible this will effect the final results.

    Another experiment try reducing the resolution by changing the following parameters.

    LCr4500_3D_Scanner_Application_v2\config\algorithm_horizontal.txt 

    GRAY_CODE_PARAMETERS_SEQUENCE_COUNT = 9 // Change this to 8 or 7 

    Similarly 

    LCr4500_3D_Scanner_Application_v2\config\algorithm_vertical.txt

    GRAY_CODE_PARAMETERS_SEQUENCE_COUNT = 9 // Change this to 8 or 7

    Note, after changing this value, you have to redo from the beginning.

    Regards,

    Sanjeev

  • 0 in reply to Sanjeev
    Intellectual 885 points
    Hi Sanjeev,

    Thanks for your reply.

    I made those changes but the point cloud is still poor for a flat wall.

    In relation to the camera have I just got it too far back from the calibration grid or is it just not good enough to get decent results with regards resolution. It is720P so not sure why it's only showing 640x480 resolution. (I'll look into that)

    Should I be able to get OK results from a camera like this? What resolution did you guys test with? We have a higher resolution camera on order but this will not be with us for 2 weeks so we are trying to better understand the physical setup required at this stage

    If my image contains squares that are 8mm in each dimension, and taking into account the images you have seen , does the setup in the picture look OK. Is there anything else you would change?

    Thanks

    JT
  • 0 in reply to John Traynor
    TI__Mastermind 30875 points
    JT,
    Okay, thanks for sharing the results.
    We use openCV camera capture functions USB cameras. Industrial cameras like Point Grey camera they have dedicated library support, in case of openCV camera we weren't able to adjust the camera resolution. You may want to investigate the openCV camera for configuration.
    You can definitely try Pointr Grey camera documented, this will be give better results, these will be global shutter and hardware triggered you will be able to run much faster scan.
    Regards,
    Sanjeev
  • 0 in reply to Sanjeev
    Intellectual 885 points
    Hi Sanjeev,

    Again thanks for the reply.

    We are getting in an industrial camera but it's not point grey. it's a dalsa genie

    What I was hoping to do was having a working setup with a low cost camera like the Microsoft Life cameras through opencv and then when our industrial camera arrives in, switch over to that. We'll have to write the SW to support this new camera so it would be good to have a working setup, albeit not the most accurate, to compare it against. What we are getting at the moment is obviously not write and we are wondering are we doing something fundamentally wrong

    If you had a picture or more details on how you integrated the 4500 EVM module with the lifecam it would help us greatly (and I'm sure other forum members who are starting off). I know there is some of this in the 00-254 document but it is limited.

    Thanks in advance

    JT
  • 0 in reply to John Traynor
    TI__Mastermind 30875 points

    JT,

    We haven't experimented LightCrafter 4500 with Lifecam camera as such. However we did some test using LightCrafter kit, attaching the scan results we got at that time. 

    I am still thinking reducing the number of graycode bits should help you get better results, please try number of levels to 7 also try reducing the projector exposure a bit see if it helps.

    LCr4500_3D_Scanner_Application_v2\config\config_projector.txt 

    I see you are using 60ms you can reduce it.

    One way to is look at the captured images, and the last scan pattern captured where we have alternate black and white lines it is visible clearly. This should help.

    Regards,
    Sanjeev

    /cfs-file/__key/communityserver-discussions-components-files/94/6201.TI_5F00_DLP_5F00_3D_5F00_SDK_5F00_LightCrafterKit.zip

  • 0 in reply to Sanjeev
    Intellectual 885 points

    Hi Sanjeev,

    Again thanks for the help

    I made the changes you suggested but it's still not making a huge difference so I am thinking there is something wrong with our system calibration procedure

    One thing I am noticing is that when I do step 5 to calibrate the system it fails a lot saying projected calibration board not found when trying to acquire the images. I have to typically try and acquire the images in this step multiple times before it says they are valid.What are the typical reasons why this would be happening? (Triggering, varying exposure on camera etc???) 

    Thanks 

    John

  • 0 in reply to John Traynor
    Intellectual 885 points
    Hi Sanjeev,

    I tried the following and it's improved things dramatically but still not sure why

    I moved the 4500 module back so that I have a bigger projected area around the calibrated board

    I changed the frame rate of the camera to 15 (was 30)

    Sequence exposure and period is now set to 30000

    Gray code bits are still set to 7 for these tests

    When I select option 6 I get something that is much better - not perfect but getting much closer.

    When I select option 7 it looks terrible. Nothing at all like a flat wall



    Why would a vertical scan be much better then a horizontal.

    Thanks

    John
  • 0 in reply to John Traynor
    Intellectual 885 points

    https://e2e.ti.com/cfs-file/__key/communityserver-discussions-components-files/94/3223.LCr4500_5F00_3D_5F00_Scanner_5F00_Application_5F00_v2.7zHi Sanjeev,

    I've attached the latest files in case they are of help to you

    Thanks in advance

    John

  • 0 in reply to John Traynor
    Intellectual 885 points
    Hi Sanjeev,

    I think I might have spotted the problem. It looks like the horizontal image is inverted in the lifecam. Is there a setting in the demo application config files that can invert this or does it need to be done on the camera itself?

    Would this explain what I am seeing based upon the images in the folder I attached in the previous post?

    Thanks

    John