Geometric Calibration for Cameras with Inconsistent Imaging Capabilities.

Area of Science:

• Computer Vision
• Geometric Calibration
• Image Processing

Background:

• Traditional camera calibration methods assume uniform uncertainty for detected control points.
• Inconsistent camera imaging capabilities lead to varying uncertainties in control point localization.
• Existing methods can overfit calibration parameters to less reliable sensor areas.

Purpose of the Study:

• To develop a novel method for measuring and incorporating control point uncertainties into geometric calibration.
• To improve the accuracy and robustness of camera calibration by addressing varying point uncertainties.

Main Methods:

• Quantifying uncertainties of detected control points based on imaging characteristics.
• Developing a new optimization model for geometric calibration that utilizes measured uncertainties.
• Implementing a weighted approach to suppress influence from high-uncertainty points and enhance low-uncertainty points.

Main Results:

• The proposed method effectively measures control point uncertainties.
• The new optimization model demonstrates improved calibration performance compared to traditional methods.
• Experimental results show reduced overfitting to poorer sensor areas.

Conclusions:

• The proposed method offers a more accurate and reliable approach to camera geometric calibration.
• Accounting for varying control point uncertainties is crucial for robust calibration.
• This technique enhances the performance of calibration modules like OpenCV's.