Camera Array Calibration Using Out-of-Focus Circular Grid Target
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The calibration and registration of camera arrays are time-consuming and expensive, requiring extensive space due to the need for capturing images from various orientations and ensuring that optical target features are in focus, which complicates the process of determining intrinsic camera parameters and registering color and depth cameras effectively.
Innovation Solution
The use of an optical target positioned outside the depth of focus, featuring a grid of symmetrical shapes, allows for the determination of intrinsic camera parameters through an iterative calibration process using a cost function that evaluates the straightness of lines intersecting rows and columns associated with centroids, enabling efficient calibration and registration of camera arrays with reduced spatial and temporal requirements.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional calibration methods using planar checkerboard patterns are used, then registration accuracy can be achieved, but the process requires extensive space and time for capturing images from various orientations
Solution Approach 1:
The patent changes the geometric parameters of the optical target from a planar checkerboard pattern to a three-dimensional circular grid pattern. This parameter change allows the target to provide sufficient calibration information without requiring multiple image captures from different orientations, thereby reducing calibration time while maintaining registration accuracy.
Solution Approach 2:
The patent transitions from two-dimensional planar targets to three-dimensional spatial targets. The circular grid pattern is arranged in three-dimensional space, allowing the system to capture calibration data from a single position without needing to rotate or reposition the target, thus eliminating the time-consuming image capture process required by traditional planar methods.
2Measurement precision
If traditional calibration methods requiring multiple image captures are used, then intrinsic parameters can be determined, but the process requires extensive space for positioning
Solution Approach 1:
The patent uses a three-dimensional circular grid pattern that can be perceived and processed from a single camera position. This spatial arrangement allows the system to determine intrinsic parameters without requiring the calibration target to be positioned in multiple locations or orientations, significantly reducing the calibration space requirement compared to traditional planar checkerboard methods.
3Measurement precision
If optical target features are positioned to be in focus, then measurement precision is improved, but the depth of focus constraints complicate the calibration process
Solution Approach 1:
The patent positions the circular grid pattern at a depth that is outside the depth of focus range of the camera. By arranging the target features in three-dimensional space rather than on a planar surface at the focus plane, the system can still detect and process the pattern using computational methods, thereby avoiding the depth of focus constraints that complicate traditional calibration processes.
Data Source
AI summary
Methods for determining intrinsic parameters associated with a camera and for registering cameras within a camera array are described. In some embodiments, a camera may be calibrated using an optical target that is positioned at a depth outside the depth of focus of the camera. The optical target may include a grid of symmetrical shapes (e.g., a rectilinear grid of circles or tilted squares). The intrinsic parameters for the camera may be determined using an iterative calibration process in which a cost function is used to evaluate the straightness of lines intersecting rows and columns associated with centroids of the symmetrical shapes. In some embodiments, the registration of a color camera with a depth camera may include mapping centroids identified within a first color image captured by the color camera with corresponding centroids identified within an undistorted intensity image captured by the depth camera.