Vehicle Camera Extrinsic Calibration Using Road Patch Epipolar Geometry
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional methods for calibrating vehicle rear camera extrinsic parameters are laborious, time-consuming, and costly due to the need for precise target maintenance and controlled lighting conditions, which affect accuracy.
Innovation Solution
A method and system for single camera calibration that utilizes a moving camera to capture images of road patches, applying an epipolar model to determine extrinsic parameters without requiring a dedicated target or controlled environment, using a camera mounted on a vehicle to capture images of road patches and applying an epipolar model to derive extrinsic parameters.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a pre-defined target with known dimensions is used for calibration, then measurement accuracy can be improved, but the complexity of the calibration process and environmental requirements increase
Solution Approach 1:
The patent replaces the physical pre-defined target with a virtual target created by generating a synthetic image of a calibration pattern. This virtual target is then applied to a reference object in the scene, eliminating the need for precise physical target fabrication and maintenance while preserving measurement accuracy through computational geometry
Solution Approach 2:
The patent substitutes the mechanical target-based calibration system with a computer vision-based system. Instead of using physical targets with known dimensions that require precise manufacturing and maintenance, the system uses image processing and geometric calculations to determine extrinsic parameters, thereby reducing physical complexity
2Measurement precision
If controlled lighting conditions are provided for calibration, then measurement accuracy is improved, but the time and cost of calibration increase
Solution Approach 1:
The system enables the camera to calibrate itself by capturing images of the virtual target applied to a reference object in the natural environment. The calibration process uses the camera's own imaging capabilities and standard image processing algorithms, eliminating the need for separate controlled lighting setups and specialized calibration equipment
Solution Approach 2:
The calibration method can be performed under various lighting conditions and environments, making the calibration process universal rather than requiring specific controlled conditions. The same algorithm works whether the camera is mounted on a vehicle, drone, or other moving platform, and whether the environment is bright, dim, indoor, or outdoor
3Measurement precision
If a pre-defined target is maintained very well for calibration, then measurement accuracy is improved, but the effort and cost of calibration increase
Solution Approach 1:
The patent creates a digital copy of the calibration target as a virtual image that can be applied to any reference object visible in the camera view. This eliminates the need for precise physical target fabrication, maintenance, and handling, while the virtual target can be perfectly reproduced computationally without degradation
Solution Approach 2:
The virtual target can be regenerated computationally if needed, replacing expensive physical targets that require careful maintenance. The digital target is essentially disposable and can be recreated without cost, eliminating the need for expensive physical target fabrication and maintenance infrastructure
Data Source
AI summary
A method for determining extrinsic camera parameters includes: starting camera movement, capturing a first raw image with parallel first and second patches at a first point of time, and a second raw image with parallel third and fourth patches at a second point of time. A distance between the first and second patches and between the third and fourth patches is the same. A reference position A of a first patch image feature, a reference position C of a second patch image feature, an offset position B of the feature of the first patch in the third patch, and an offset position D of the feature of the second patch in the fourth patch are detected. An epipolar model is applied based on the positions A-D and a distance travelled by the camera between the first and second time points. Extrinsic camera parameters are determined from the epipolar model.