Method and system for determining camera parameters from a long range gradient based on alignment differences in non-point image landmarks

Number: 9883163
Published: 2018-01-30
Filed: 2013-01-09
Assignee: DISNEY ENTERPRISES, INC.
Inventors: Carr; G. Peter K.; Matthews; Iain
CPC: H04N13/207; G06T7/85
Verdict: Low Notable software
Source: Google Patents · FreePatentsOnline

The keeper's note

Camera-parameter determination from long-range gradient.

Abstract

A system and method improves a camera calibration. The method includes receiving a camera image, a planar template pattern, a 3D geometry of a surface on which the planar template pattern is embedded, and a set of parameter values. The method includes rendering the planar template pattern into a camera perspective based on the parameter values to generate a warped template image. The method includes generating an error image including at least one non-zero difference between the camera image and the warped template image. The method includes adjusting the parameter values to reduce an error between the camera image and the warped template image.

Background

BACKGROUND INFORMATION(1) Camera calibration has been done on the basis of point-based targets like checkerboards, but such targets are not practical for outdoor-based cameras, such as those used in outdoor sporting environments. Standard camera calibration algorithms employ a two-stage approach that first identifies point correspondences between geometric features in an image and in the world (e.g., a checkerboard's junction points from its camera's image projection and from the real world positions), and then finds the perspective transformation parameters which minimize the distance between the transformed world points and their corresponding image locations. Checkerboards are perhaps the most popular target because of their ease of use and manufacture, but in large outdoor scenes, checkerboards are impractical, as the necessary size may be on the order of meters. Furthermore, detecting a sufficient number of landmark point features continuously from all camera field-of-views may be challenging given the highly dynamic steering of a broadcast camera (especially when covering a sporting event where camera panning and zooming may be rapid and extreme). What is needed is a camera calibration technique that relies on non-point landmarks in a scene, also known as an image-based alignment. Contrary to a feature-based alignment method, for example, an image-based alignment method aligns non-point landmarks in a reference image with counterpart non-point landmarks in the scene ima

Claims

1. A method, comprising: receiving a camera image captured by a camera and a set of parameter values that models a perspective of the camera; receiving a synthetically generated planar template pattern embedded on a surface and a 3D geometry of the surface, wherein the planar template pattern is a representation of a sporting field; calculating a long range gradient of the planar template pattern, wherein the long range gradient of the planar template pattern is based on a calculation within a gradient window; warping the planar template pattern into the perspective of the camera based on the set of the parameter values to generate a warped template image; generating an error image based on a comparison between the camera image and the warped template image; and computing an adjusted set of parameter values to reduce an error associated with the error image, wherein the computing is based on the long range gradient. 15. A system, comprising: an image capturing device configured to generate a camera image; a memory arrangement configured to store the camera image, a synthetically generated planar template pattern that is a representation of a sporting field embedded on a surface, a 3D geometry of the surface, and a set of parameter values that models a perspective of the image capturing device; and a processor configured to warp the planar template pattern into the perspective of the image capturing device based on the set of parameter values to generate a warped template image, the processor configured to calculate a long range gradient of the planar template pattern, wherein the long range gradient of the planar template pattern is based on a calculation within a gradient window, the processor configured to generate an error image based on a comparison between the camera image and the warped template image, the processor configured to compute an adjusted set of parameter values to reduce an error associated with the error image, wherein the computing is based on the long range gradient.