Logging reprojection error after static camera calibration

Autonomous Vehicles DRIVE AGX Xavier DRIVE AGX Xavier General
, ,
1

Hardware Platform: DRIVE AGX Xavier™ Developer Kit
Software Version: DRIVE Software 10
Running on the platform
Host Machine Version: Ubuntu 18.04

Hi,
I have successfully perform extrinsic calibration using the Graph calibration tool. I have also run the Calibrated Graph to rig tool and generated a rig.json file.

The Graph calibration tool outputs the following type of information for the different external images and cameras involved in the calibration (have attached sample output below)

[30-09-2020 15:51:23] Saved calibrated graph to ./calibrated-graph.json
[30-09-2020 15:51:23] Writing validation images
Errors for intrinsics external-IMG_0924.JPG: Median=0.404315, mean=0.395903, var=0.0367015, std dev=0.191576, sample count=32, min=0.0935775, max=0.952394
Errors for intrinsics external-IMG_0924.JPG: Median=0.777376, mean=0.759124, var=0.061046, std dev=0.247075, sample count=140, min=0.153108, max=1.45862

Q1: Are there any options (command line options with ./calibration-graph-cli (or) ./calibration-graph-to-rig) which allows me to log all the reprojection error information (in say a .json file) for different cameras and externals images?
Notes: I have analyzed the different artefacts from the calibration process and I don’t seem to see the reprojection error being logged anywhere (except the command line when I run the tool obv.)

Q2: Other than the validation images generated is there any other ways to validate the extrinsics in Driveworks (like the intrinsics validator tool)?

Thanks in advance!

3

Dear @asaraiya,
Q1. No.
Q2. The generated extrinsic validation images actually correspond to the intrinsic validation images of the intrinsic validation tool, as they both depend on AprilTags as an input for the validation. Still, a different option for extrinsic validation is to run the DW camera self-calibration using the static calibration results as the nominals in the rig, and verify that the estimated corrections are small (e.g., using sample_calibration_camera or directly using the dwCalibrationEngine API

4

Dear @asaraiya,
Do you have any questions on related topic?