.png){kind=icon}
.png){kind=icon}
Return
Wide-Range Camera Calibration Based on Normalized Localization Variance With Applications
.png){kind=icon}
PRE
AI
DOI：10.1109/TIM.2025.3575993.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
The Gauss-Markov theorem states that achieving an optimal solution requires homoscedasticity of error terms. However, during camera calibration, target placement affects image clarity and feature scale, leading to varying localization variances and suboptimal calibration results. Existing methods assume uniform localization variance within the depth of field, but this assumption breaks down in wide-range measurement scenarios where the target moves beyond the depth of field, degrading calibration accuracy and measurement precision. To address this issue, a wide-range camera calibration method based on normalized localization variance is proposed. A localization variance model for checkerboard corner is established, defining the relationship between image processing scale and localization variance, from which the normalized localization variance is derived. This normalized localization variance is incorporated into the nonlinear optimization of calibration parameters to ensure compliance with the Gauss-Markov theorem and achieve theoretical optimality. Experimental results validate the effectiveness of the proposed method, demonstrating reduced measurement errors and improved calibration accuracy across various applications, including satellite reconstruction and pose measurement application.
Keywords:
Feature scale normalization
Gauss-Markov theorem
localization variance model
vision measurement
wide-range calibration
Journal
IF:
5.9
Papers:
1.9W
Citations:
5.8W