Method and system for calibrating multiple cameras
AssigneeStarship Technologies Oü
InventorsRasmus RAAG, Indrek MANDRE
Life of this patent
Estimating the term…
Abstract
The invention relates to a method for calibrating cameras. In particular, the invention provides a method that estimates the position of interest points such as corners on images of array patterns, and uses such refined positions in an iterative calibration algorithm. The method can be extended to simultaneously process information from a large number of images from a plurality of cameras that are arranged in a fixed position on a mobile robot, wherein the images are representative of different camera poses. The optimization thereby provides a highly accurate estimate of camera calibration parameters.
Claims24
A method of calibrating at least one camera, the method comprising:
- (a) capturing at least one image of at least one array pattern, said at least one image comprising a plurality of interest points, with at least one camera for at least one pose of each of the at least one camera;
- (b) processing said at least one image to determine a provisional position of at least one interest point on the at least one image;
- (c) refining the provisional position of said at least one interest point on the at least one image by:
- (i) selecting a partial area of the at least one image that includes at least one interest point and at least a portion of at least two edges that meet at or near the interest point,
- (ii) defining at least a first set of edge points and a second set of edge points representative of said at least two edges;
- (iii) applying regression analysis to the first set of edge points and the second set of edge points, respectively;
- (iv) determining a refined position of the at least one interest point as an intersection of regression functions representative of the first of edge points and the second set of edge points;
- and
- (d) optimizing camera calibration parameters based on the refined position of the at least one interest point on the at least one image.
An image capture system comprising:
- (a) at least one land-based mobile robot;
- (b) at least two cameras that are fixedly mounted on each said at least one mobile robot;
- (c) a processing module, adapted to:
- (i) receive and store image data obtained by the at least two cameras of at least one array pattern that comprises a plurality of interest points;
- (ii) process said image data to determine a provisional position of the interest points on each captured image;
- (iii) refine the provisional position of the interest points by regression analysis of data points representative of at least two edges that meet at or near the interest points, the refined position being determined as an intersection of regression curves thereby obtained;
- and
- (iv) optimize camera calibration parameters based on the refined position of the interest points.
A method of detecting interest points on an image of an array pattern, the method comprising:
- (a) capturing at least one image of an array pattern comprising a plurality of interest points with at least one camera;
- (b) processing said at least one image to determine a provisional position of at least one interest point on the at least one image;
- and
- (c) refining the provisional position of said at least one interest point on the at least one image by:
- (i) selecting a partial area of the image that includes at least one interest point and at least a portion of at least two edges that meet at or near the provisional position of the interest point,
- (ii) defining at least a first set of edge points and second set of edge points representative of said at least two edges;
- (iii) applying regression analysis to the first set of edge points and the second set of edge points, respectively;
- and
- (iv) determining a refined position of the at least one interest point as an intersection of regression functions representative of the first set of edge points and the second set of edge points.
A method of calibration of at least one camera, the method comprising:
- (a) providing a mobile robot comprising a plurality of cameras that are fixedly mounted thereon;
- (b) providing at least two geometrically fixed array patterns, each pattern comprising a plurality of interest points;
- (c) capturing a plurality of images of at least a portion of said array patterns using said plurality of cameras for a plurality of poses of the mobile robot;
- (d) determining the position of a plurality of interest points within each of the array patterns for images captured in (c), so as to generate a grid of interest points within each array pattern for different poses of the mobile robot;
- and
- (e) process image information from (d) to simultaneously optimize at least one of:
- (i) position of said plurality of cameras;
- (ii) optical center of each of said plurality of cameras;
- and
- (iii) focal length of each of said plurality of cameras.
Dependencies are inferred from the claim text.