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.

The image captured by FL3-U3-88S2C-C Camera is half white and half dark when I perform calibration, why?

Other Parts Discussed in Thread: TIDA-00254, TIDA-00361

  • Hi Chester,

    It may be due to the fact you are using a rolling shutter color camera. Our currently released TI design is setup for use with the monochrome global shutter version of the PG Flea3. We will be releasing a new version of the DLP SDK by the end of March that will support the color rolling shutter version of the Flea3.

    In the meantime, you could take a look at the source code to see what needs changing or use the monochrome global shutter variant if you have that option available.

  • Just wanted to chime in:

    Is there any remaining issue using the Flea3 (FL3-U3-32S2C-CS)? I'm using the latest TIDA-00254 3D Scan Application and going by the settings in the documentation that is suggested:

    - rolling shutter color camera, make sure PG_FLYCAP_PARAMETERS_PIXEL_FORMAT is set to "MONO8."

    - PG_FLYCAP_PARAMETERS_STROBE_DELAY to "5.0" for rolling

    - projector sequence exposure be more than 50 ms


    My calibration seems to work (error ~1 and images look good with uniform lighting) but when I actually go to scan an object or a wall the scan is just garbage. Is there another setting that I need for this Flea3 model? Thanks.

  • Hi Clifford,

    What type of scan are you running and how is your set-up configured? The reason I ask is that if you have the camera mounted directly above the Lightcrafter and you do a vertical only scan, you won't get much information and the scan will look bad.

    I would also look at your scan images and make sure that you have enough brightness in the scene and that everything is in focus.

    If you're able to calibrate and capture images, it seems unlikely that the camera would be causing any issues. Looking at the scan images though will be revealing.

  • Thanks for the reply. The camera is located approximately 12" to the left of the projector, slight higher (2-3" higher). I have attached a picture.

    I think the brightness is good as you can see from the picture above. Although, I would assume that most of the lighting would come from the projector. Also, I'm fairly certain all is in focus as I acquired the scan right after calibration and nothing was changed and the cal was good (~1.0-1.5 error for both camera and projector). Calibration was performed at almost the same distance as from the brick wall (minus the size of the calibration board). The distance to the wall is ~32". Lens is a fujinon 6mm/1.2:1 focus with locks (locked tight after cal). 

    I have attached a picture of the scan (point cloud) to show as well. It is hard to see, but the points are almost in a conical shape emanating from the big cluster on the left side. Most of the scans I have taken result like this. I have done this process approximately 20 or so times (cal camera + projector, then scan a flat surface as outlined in the docs). Occasionally, I will get a scan where half  the points look like the object I'm scanning and the other half is just garbage (not sure if that detail helps). The can is a full scan (vertical + horizontal).

    Any help would be greatly appreciated. Let me know if any other details are needed. Thanks.

  • Clifford,

    Your setup looks great. I think I may not have been clear about my brightness comment - the ambient light should not be particularly bright, but the scene as it is illuminated by the projector should provide most of the light, as you noted.

    When you look at the output images (captured images) do they look good? They should have high contrast and the whole pattern should be captured.

    Are the results the same regardless if you do Vertical Only, Horizontal Only, or Both?

  • Thanks for the followup.

    I have previously attempted the calibration and scanning in total darkness with the same result (about 5 times). So it seems independent of ambient light. The scan I attached was done in an office with no outside sunlight. So some ambient light but the projector was definitely bright enough to over whelm the ambient light.

    My expectation of the points is that they should look similar to that of the "Calibration Verification" scan in the documentation (Section 3.8 from "TIDA-00254 -Accurate Point Cloud Generation", filename dlpu019a.pdf). Essentially it should be a flat surfaced wall (actually the same wall in front of the setup in the previous post image). So the scan is pretty far off.

    As for your suggestion of vertical and horizontal scan separately. I just tried it and vertical scan pattern produced no points. The horizontal scan produced essentially the same as the full scan. So apparently, the points from the full scan are just from the horizontal scan.

    I was also wondering about the calibration board (see image above). I included it in case something jumps out as a problem. I used tape to place it on a poster board. There are slight wrinkles but not bad. The square size is 21 mm. The poster board is matte, so no shiny material. It is placed approximately where I do the cals.


  • During my investigations, I came across some additional information (may or may not be an issue). I mentioned earlier, that sometimes only half the scan is bad (usually the top half). I was looking I recently scanned again and while browsing through the "output/scan_images" directory (where the captured pattern images are stored), I notice that some (not all) have a significant blur horizontally in the center. This location corresponds to parts of the resulting scanned image where there seems to be problems. The image below shows both images (left pattern projected image 13, right is the color map from the full scan). Ideas? Thanks for your help in advance.

  • From the image you attached above, what looks like is happening is that you're catching the transition between patterns with the rolling shutter. The top half looks offset by one line to the bottom half (at least to my eyes).

    Aside from modifying the config files to account for the rolling shutter, have you had to make any other adjustments?

    Let me do some research and see what the SDK does to handle the rolling shutter. If you could maybe post a couple more output images so I can see any other abnormalities, that would be great.

  • Sure. Thanks.

    Here are five more.

  • Sure looks like you're having issues with syncing and the rolling shutter. I'll take a look at how the system handles the rolling shutter and give you some suggestions.

    If you have a global shutter camera around, give that a shot and see if you have better results.

  • Clifford,

    Can you confirm that your config files match mine? The projector settings need to be mirrored to the camera settings for this to work properly as well.



    ##OpenCV Camera    #
    ##PointGrey Camera #
    # For below parameter set 
    # Rolling shutter color Camera - MONO8
    # Global shutter Monochrome camera - RAW8
    # For below parameter
    # Rolling shutter color camera - 5.0
    # Glocal shutter Monochrome camera - 0.0

    ##For below parameter
    ## 1 - External Positive Trigger
    ## 0 - Internal Trigger
    LCR4500_PARAMETERS_DLPC350_FIRMWARE		= C:\Texas Instruments-DLP\DLPR350PROM-3.0.0\DLPR350PROM_v3.0.0.bin

  • Yes, it is exactly.
  • Thank you for looking in to this. I can confirm that with a camera that utilizes a global shutter, that the camera and projector calibration work and I'm able to get reasonable scans.
  • Okay. We have a rolling shutter color camera here and I'm going to give it a shot and see if there is something else that needs to be configured. I'll post back once I've attempted a few scans.

  • I was able to successfully make a scan with my camera. We have a  FL3-U3-12S2C-CS here and it worked fine.

    Could you do the following? Open the Point Grey Flycapture2 software and click on the circled button. If you could show us what your trigger settings and advanced camera settings are, we can compare and see if anything else looks like it needs tweaking.

  • Here are the trigger and advanced settings for the Flea3 U3-32S2C:

  • Wow, I'm stumped...let me ask some other folks here who were involved in the implementation of rolling shutter with the SDK and see if they have any thoughts.

  • Clifford,

    Two things:

    1. Can you lower your camera exposure but leave the projector exposure? Try a scan with these settings.

    2. When you do a scan, in the ...\TIDA-00361_3D_Scanner_LCr3000-2.0\LCr3000_3D_Scanner_Application_v2\output\scan_images folder, are there black frames at the end of the sequence?

    The results from this will help us narrow whether it's an exposure issue or a trigger issue.