Non-Central Camera Calibration for Wide-Field Mapping Accuracy
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing camera calibration methods, particularly for wide-field of view cameras like fisheye cameras, assume a central camera model with a pinhole aperture, leading to errors due to the assumption of a point-like aperture, which is not valid for non-central cameras with larger apertures.
Innovation Solution
A non-central camera model is used that accounts for a larger aperture, incorporating a distortion transform and viewpoint offset to accurately map object points to image points, involving a series of transforms including intrinsic and extrinsic transforms, and a parametric function to model the offset along the optical axis.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If a central camera model with pinhole aperture is used for calibration, then the calibration process is simple, but the mapping precision deteriorates for non-central cameras with larger apertures
Solution Approach 1:
The patent changes the fundamental parameters of the camera model from a pinhole aperture assumption to a non-central aperture model with a larger entrance pupil. This involves modifying the projection equations to include a non-zero offset between the optical center and the aperture center, thereby accurately representing the actual camera geometry and improving mapping precision for wide-field cameras.
Solution Approach 2:
Instead of assuming a simple pinhole model and correcting for its limitations, the patent inverts the approach by directly modeling the actual non-central aperture geometry from the beginning. The calibration process explicitly parameters the entrance pupil position and size, and the projection equations are formulated to account for this non-central aperture configuration, eliminating the need for subsequent corrections.
2Measurement precision
If a non-central camera model with larger aperture is used, then the mapping accuracy improves, but the calibration complexity increases
Solution Approach 1:
The patent segments the calibration process into distinct modules: capturing calibration images, detecting calibration pattern features, determining correspondence between object and image points, and optimizing the projection model parameters. This segmentation allows the complex non-central camera calibration to be broken down into manageable steps, each with clear objectives and computational tasks.
Solution Approach 2:
The calibration process incorporates feedback through iterative optimization. The initial projection model parameters are used to generate predicted image points, which are then compared with actual detected points. The difference (error) is fed back to adjust and refine the parameters, progressively improving the mapping accuracy until convergence is achieved.
Data Source
Figure 1~4B
AI summary
An apparatus that calibrates a parametric mapping that maps between object points and image points. The apparatus captures an image of a calibration pattern comprising features defining object points. The apparatus determines, from the image, measured image points that correspond to the object points. The apparatus determines, from the mapping, putative image points that correspond to the object points. The apparatus minimizes a cumulative cost function dependent upon differences between the measured image points and putative image points to determine parameters of the parametric mapping. The mapping uses a parametric function to specify points where light rays travelling from object points to image points cross the optical axis.