How To Get Good Quality Data On Zivid Calibration Board

This tutorial aims to present how to acquire good quality point clouds of the Zivid calibration board for hand-eye calibration. It is a crucial step to get the hand-eye calibration algorithm to work as well as to achieve the desired accuracy. The goal is to configure camera settings that provide high-quality point clouds regardless of where the calibration board is seen in the working space.

It is assumed that you have already specified the robot poses at which you want to take images of the calibration board. In the next article, you can learn how to select appropriate poses for hand-eye calibration.

Note

To calibrate using the Zivid calibration board, ensure that the entire board, including the fiducial marker, is fully visible for each pose.

We will discuss two poses: the ‘near’ pose and the ‘far’ pose. The ‘near’ pose refers to the robot position where the imaging distance between the camera and the checkerboard is minimized. In eye-in-hand systems, this is when the robot-mounted camera is closest to the checkerboard. In eye-to-hand systems, it is when the robot positions the checkerboard closest to the stationary camera.


Eye-in-hand robot pose for a close capture of Zivid calibration board	Eye-to-hand robot pose for a close capture of Zivid calibration board

The ‘far’ pose refers to the robot position with the greatest imaging distance between the camera and the checkerboard. In eye-in-hand systems, this is when the robot-mounted camera is farthest away from the checkerboard. In eye-to-hand systems, it is when the robot positions the checkerboard farthest away from the stationary camera.


Eye-in-hand robot pose for a close capture of Zivid calibration board	Eye-to-hand robot pose for the farthest capture of Zivid calibration board

Tip

Try Calibration Board presets for the nearest and farthest away pose. If you get a good result, you can skip the rest of the tutorial and use these settings.

The expected result is illustrated below:

../../../_images/how-to-get-good-quality-calibration-board-final-result.png

A step-by-step process for acquiring good point clouds for hand-eye calibration is presented as follows.

Possible scenarios

In the following tutorial, we will use the SNR map to check the signal quality of black and white pixels. The color indicator changes from red to dark blue as the SNR value increases. Below, you will find the SNR scale and all the possible cases you may come across in this tutorial.

../../../_images/how-to-get-good-quality-calibration-board-horizontal-SNR-scale.png


The black areas are underexposed, and the quality of the white areas is not good.	The quality of the black areas is not good and the white areas have a satisfactory quality.	The quality of the black areas is satisfactory and the white areas have an optimal quality.	The quality of the black areas is optimal and the white areas are overexposed.	The black and white areas are overexposed.

Base settings

First we will define the base settings for this tutorial.

Move the robot to the ‘near’ pose
Start Zivid Studio and connect to the camera
Set Vision Engine to phase
Set Sampling, Color to rgb.
Set Sampling, Pixel to all.
Set Exposure Time to 10000μs for a 50Hz grid frequency or 8333μs for a 60Hz grid frequency
Set f-number using depth of focus calculator
- Minimum Depth-of-Focus (mm): Farthest working distance - Closest working distance
- Closest working distance (mm): is the closest distance between the camera and the checkerboard
- Farthest working distance (mm): is the farthest distance between the camera and checkerboard
- Acceptable blur radius (pixels): 1
Set Projector Brightness to maximum
Set Gain to 1
Set Noise Filter to 5
Set Outlier Filter to 10
Set Reflection Filter to global

Turn off all other filters and leave all other settings to their default value.

Optimizing camera settings for the ‘near’ pose

Fine-tuning for ‘near’ White (Acquisition 1)

At this stage, ignore the black checkers and focus on the white checkers. Capture and analyze the white regions of Zivid calibration board. Use the following images as an example when Fine-tuning for white regions.


The quality of the the white areas is optimal, the black areas can be ignored (SNR map).	The quality of the the white areas is optimal, the black areas can be ignored (Depth map).

At this point, one of four acquisitions (“Acquisition 1”) is tuned.

Fine-tuning for ‘near’ Black (Acquisition 2)

Turn off the “Acquisition 1” and clone it. This “Acquisition 2” needs to be tuned to have good data on the back part of Zivid calibration board. Therefore, it’s expected to overexpose the white checkers. Look at the following image as an example, where we have no data on the white part of the checkerboard due to overexposed white pixels:


White pixels overexposed and good data on the black pixels (SNR map).	White pixels overexposed and good data on the black pixels (Depth map).

Darker surfaces require higher light exposure.

Optimizing camera settings for the ‘far’ pose

Fine tuning for ‘far’ White (Acquisition 3)

Turn off the “Acquisition 2” and clone the “Acquisition 1”. Again, let’s ignore the black checkers and just focus on the white ones. Capture and analyze the white regions of Zivid calibration board.

Fine-tuning for ‘far’ Black (Acquisition 4)

Turn off the “Acquisition 3” and clone the “Acquisition 2”.

This “Acquisition 4” needs to be tuned in order to have good data on the back part of Zivid calibration board. Therefore, it’s expected to overexpose the white checkers.

Darker surfaces require higher light exposure.

At this point the settings for the four acquisitions are configured. Now it’s just missing adding the final filters.

Optimizing filters

Set Gaussian Smoothing to 5
Set Contrast Distortion, Correction to 0.4
Set Contrast Distortion, Removal to 0.5

Your point cloud should be similar to the point cloud shown at the top of the tutorial.

Note

In most cases, two acquisitions, one optimized for ‘near’ and the other for ‘far’ poses, will provide a good quality data on the checkerboard. Another option for determining the correct imaging settings is to utilize the Capture Assistant. However, it currently only offers optimal settings for the entire scene, rather than specifically for the checkerboard. While the Capture Assistant can still be effective for this purpose, it is recommended to use the method described above, as it consistently yields good results.

Let’s see how to realize the Hand-Eye Calibration Process.