DLPLCR4500EVM: system calibration (camera + projector)

Hi Renan,

Please provide the setup procedure including commands sent to observe these artifacts. Also, please provide images of the distortions.

Regards,

Austin

Hello Austin,

Please see the update from my customer:

I have to clarify one thing, I am working right now with the DLP projector DLP3010EVM (TI product, adapted from EKB Techn.).

• Projector: DLP® 3010 Light Control Evaluation Module (ekbtechnologies.com)
• Cameras: Hikrobot - Machine Vision - Product Detail (hikrobotics.com)

I adapted the SL-SDK and the machine vision software slightly, to allow calibration and reconstruction.

I am 99% that this all works like it should.

My only problem is the system-calibration of projector + camera...!

In the download link (https://www.swisstransfer.com/d/3ce40de2-0bdf-4dde-8883-add0e151b7ff), I send you 4 different system calibration tests including all images, calibration files and 3D data outputs.

I would be very happy to improve the system calibration procedure, i.e., because they cost me a lot of time until it works.

Regards,

Renan

Hello Renan,

Thanks for the additional information for the customer we will need some time to look into this issue.

Thanks,

Alex Chan

Hello Renan,

To confirm, your client is using a DLP3010EVM-LC and not a DLPLCR4500EVM. Is that correct?

What is the request here? Optimizing for a third party camera is outside of our normal scope of support.

Regards,

Austin

Hello Austin,

Please see response from my customer:

Yes, I am using the DLP 3010 from EKB Technologies (same base as Texas Instruments DLP3010EVM-LC).

But I had the very same issues also with the DLPLCR4500EVM:

1. Camera calibration very easy
2. System calibration very difficult. Many times large reprojection errors with bad 3D reconstruction.

Regards,

Renan

Hello Renan,

The team will have to look at the compatibility of the SDK and the DLP3010EVM. This tool kit was meant for the DLPDLCR4500EVM.

The DLP3010EVM-LC should have its own API available on its product page as well.

Regards,

John

Hello Renan,

Sorry for the delayed response was trying to discuss the question with the team more.

In theory the SDK for the DLP4500 system could be extensible to other chipsets and projectors but there is no current support for this.

We believe the issue could center around mechanical stability of the projector, screen, and camera for each image during the calibration process as well as the screen flatness. 

When there are inconsistencies in the captured images due to changes in angles or position of either the projector, screen, or camera, there may be errors in the homography. 

Given the customer saw issues with the system calibration, can the customer check the projector and camera are mechanically stable to each other during the calibration process?

Thanks,

Alex Chan

Hello Alex,

Please see the response from my customer:

Thank you very much for the info and the support.

The projector is mounted, together with the camera(s) into Tough-PLA 3D printed frame. It’s not perfectly stable, like a metal case would be. This surely has it’s influence.

When you mention the flatness of the chessboard, I see that this is also not perfectly flat. Also to be improved.

Regarding the acquired calibration photos I have sent you, do you see there any issues? E.g., bad camera exposure time, not enough or too much ambient light, etc.?

 And maybe I mention again, I could successfully calibrate the system now (projector DLP3010EVM-LC + left and right camera). The problem is that I don’t know why it works only one of ten times.

 Regards,

Renan

Hello Renan, 

We will need some more time to look into this issue further.

Thanks,

Alex Chan

Hello Alex,

Any update?

Regards,

Renan

Hello Renan,

I am inquiring internally with one of our system engineers and hope to get a response back soon. Thank you for your patience.

Thanks,

Alex Chan

Hello Alex,

Any update?

Regards,

Renan

Hello Alex,

Good day.

Please see response from my customer:

I don’t think it’s the camera calibration. It’s always successful, I move the 3D scanner system and the chessboard is at the wall. I make images from different angles, and positions (vertical and horizontal shift by hight-adjustable table). Reprojection error always very small. And, yes, I do 20 images for the camera calibration.

Like I mentioned, I have problems with the system calibration. There I do 5 images. Also, from different angles and positions. I also tried more images, with no better results.

I think for the future, I should test a stiffer mounting option off the camera and projector (no 3D print, but metal). And using a chessboard, which is glued on really flat surface (now it is on a flat wall, but maybe not perfect).

Regards,

Renan

Hello Renan,

The team will look over the client's feedback.

Thank you,

John