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.

SK-TDA4VM: Manual Calib tool

Part Number: SK-TDA4VM

Hi team,

Here's an issue from the customer may need your help:

The customer is doing a surround view image (four way) stitching.

It's find that if they want to complete the stitching of four images,  LENS.bin, CHARTPOS.bin, and CALMAT.bin these 3 files are required. And then compile and produce a VX_app_srv_fileIO.out file using VX_app_srv_fileIO and with this .out file, the stitching of four images can be done.

They generate LENS.bin (internal parameter) through the Manual Calib tool (i.e. /vision_apps/tools/3D_calibration_tool). However, since the scaling method is different (the customer got the scaling picture from front to rear and left to right, but the Manual Calib tool is four corners), chartPOS.bin and CALMAT.bin cannot be generated.

The code that the consumer wants to call through the Manual Calib tool will modify the input to generate the CALMAT.bin file to their own external parameters. However, since the Manual Calib tool was opened directly with a main.exe , its calling code could not be found. Is there any way to fix it OR where can they get the code?

The first diagram below is the Manual Calib tool scaling diagram, while the customer scaling images are front, rear, and left, right:

Could you help check this case? Thanks.

Best Regards,

Cherry

  • Can you elaborate more on the followings? I cannot understand clearly.

    1. However, since the scaling method is different (the customer got the scaling picture from front to rear and left to right, but the Manual Calib tool is four corners)

    2. The first diagram below is the Manual Calib tool scaling diagram, while the customer scaling images are front, rear, and left, right:

     What do you mean by scaling images?

    Best regards,

    Do-Kyoung

  • Hi Do-Kyoung,

    Thanks for your support.

     What do you mean by scaling images?

    It's calibration images.

    And please let me specify the case:

    1) Installed MCR8.4 as referenced the /vision_apps/tools/3D_calibration_tool/PSDK_RTOS_UserGuide_3D_SurroundView_Manual_CalibTool.pdf doc.

    2) Then click /vision_apps/tools/3D_calibration_tool/main.exe to open the TI Surround Vision Calibration Tool. The tool generates the necessary files for three surround view image stitching in four steps: LENS.bin, CHARTPOS.bin, CALMAT.bin. 

    3) Since the camera model and scene needs are different, customers can get LENS.bin using the TI Surround Vision Calibration Tool, while these two files, CHARTPOS.bin and CALMAT.bin, are not available directly from their calibration images. 

    4) So the case here is how to use the customer's own calibration image (different placement than the official demo, where the calibration plate is placed at four corners of the vehicle, while their calibration plate is placed at the front, rear, right, left ) to get the CHARTPOS.bin, CALMAT.bin files that can be used? 

    Thanks and regards,

    Cherry

  • Okay. I now understand. The customer wants to place charts (calibration plate) in different positions. Unfortunately, this tool supports only four charts placed at four corners of a vehicle. Do you know why the customer wants to change the chart positions? Is it because of camera positions?

  • Hi Do-Kyoung,

    Do you know why the customer wants to change the chart positions? Is it because of camera positions?

    Since they just discovered this TI Surround Vision Calibration Tool after calibration is completed.

    The current idea is to get a rotating matrix between four images, regardless of the position of the calibration. They can substitute the results of the official demo with the rotating matrix between the images that are currently being calibrated in such a way, and then get the two files, ChartPOS.bin and CALMAT.bin. 

    The question is, how to replace this result of the rotating matrix between four images? 

    Thanks and Regards,

    Cherry

  • I cannot give right answers because I don't know the details of the customer's setup.

    This tool outputs M_gci = [R|t]_gci of the i-th camera w.r.t. to ground (calibraiton chart). For example, M_gc1 means [R|t] of the i-th camera w.r.t. ground. So, in order to get M_c1c2 (i.e., [R|t] from the 1st camera to the 2nd camera), 

    M_c1c2 = inv(M_gc1)*M_gc2 = M_c1g * M_gc2 

    Hope it helps.