# Interpretation of camera intrinsic parameter check results and extrinsic parameter data

## Purpose

To provide camera users with a clearer understanding of the results of intrinsic parameter check and the extrinsic parameter files, we provide the following detailed explanation.

# Interpretation of Camera Intrinsic Parameter Check Results

## Where is the tool

Tools → Intrinsic Parameter Tool → Check intrinsic parameters

Note: For the Laser L Enhanced camera, when inspecting intrinsic parameters using Mech-Eye Viewer, please uncheck "3D Downsampling.

## Interpretation

Here is an example of result for monocular camera (this example meets the standard):

• Scalar difference: 0.02 mm / 80.00 mm
• Vector difference: 0.24% 0.17 mm / 70.00 mm
• Mean relative positional error: 0.13 mm

Here is an example of result for binocular camera (this example does not meet the standard):

• Scalar difference: 0.15 mm / 100.00 mm
• Vector difference: 0.34% 0.34 mm / 100.00 mm
• Mean relative positional error: 0.30 mm
• Mean alignment error: 2.48 px

Green font indicates that the current intrinsic parameters meet the standard, while red font indicates that the current intrinsic parameters do not meet the requirements.

The result involves four values: scalar difference, vector difference, mean relative positional error, and mean alignment error.

In the image below:

1. Ground truth of distance
2. Measured value of distance
3. The magnitude of the vector represents the vector difference

The circles in the image above represent the circles on the calibration board.

Scalar difference: Refers to the difference between the measured distance between the centers of two circles on the calibration board and the ground true distance. Scalar difference can represent a certain absolute error and is expressed as a percentage.

Vector difference: Refers to the subtraction of two points in space in vector form, as shown in the formula in the image. The magnitude of the resulting vector difference represents the numerical value of the vector difference. It is measured in millimeters.

Mean relative positional error: Refers to the average deviation in positions between multiple points (designated as ‘n’) on the calibration board and the actual measured positions of these points (designated as ‘N’). After fitting the large ‘N’ and small ‘n’, obtian the mean value of the deviations in multiple positions. Measured in millimeters.

Mean alignment error: Refers to the epipolar line constraint error in a binocular camera setup, commonly understood as the relative positional relationship between the two 2D cameras of a binocular system. It is represented by calculating the distance from each circle’s center coordinates on the right camera to the corresponding epipolar line on the left camera. This error reflects the pixel difference in the relative pose of the left and right cameras compared to the theoretical values in the image. Measured in pixels.

Note: Mean alignment error only exists in binocular structured light cameras. Monocular cameras and UHP cameras do not have average alignment error.

# Interpretation of Camera Extrinsic Parameter Data

## Format of Extrinsic Parameter Sets

The format of the main extrinsic parameter set is shown in the figure, including the images captured with extrinsic parameters, calibration board parameters, extrinsic calibration configuration parameters, extrinsic parameter file, and intrinsic parameter file.

### Format of Extrinsic Parameter File

Meanings of different sections in the extrinsic parameter files.
In the screenshot:

1. Extrinsic parameters for reference frame transformation
2. Compensation parameter matrix
3. Camera resolution

### Format of Intrinsic Parameter File

Meanings of different sections in the intrinsic parameter files:
In the screenshot:

1. Distortion coefficients (if it’s 0, it means default distortion removal)
2. Camera intrinsic parameters