Calibration of a wide angle stereoscopic system

Area of Science:

• Computer Vision
• Photogrammetry
• Robotics

Background:

• Stereoscopic systems are crucial for 3D reconstruction but suffer from calibration inaccuracies.
• Errors in point correspondences and localization degrade system performance, particularly with wide-angle lenses.

Purpose of the Study:

• To develop a robust technique for correcting point locations in stereoscopic images prior to calibration.
• To mitigate errors in fundamental matrix estimation and lens distortion modeling.

Main Methods:

• A novel point correction technique is proposed, utilizing both geometrical and epipolar constraints.
• A nonlinear minimization problem is employed to refine point locations based on image neighborhood and correspondence information.
• This correction is performed before estimating the fundamental matrix and lens distortion models.

Main Results:

• The proposed method effectively avoids errors associated with point correspondences and localization.
• Improved accuracy in stereoscopic system calibration is achieved, particularly for systems using wide-angle lens cameras.

Conclusions:

• Pre-calibration point correction significantly enhances the accuracy of stereoscopic systems.
• The technique offers a robust solution for improving 3D reconstruction and calibration in challenging imaging conditions.