METHOD FOR CALIBRATION OF A STEREO CAMERA

• US 20190019309A1
• Filed: 06/01/2016
• Published: 01/17/2019
• Est. Priority Date: 08/07/2015
• Status: Active Grant

• Current Assignee: Xovis AG
• Original Assignee: Xovis AG
• Inventors: HERRLI, Markus, PAULI, Adrian
• Granted Patent: US 10,679,380 B2
• Time in Patent Office: Days
• CPC Class Codes: G06F 2218/08 Feature extraction
• CPC Class Codes: G06T 2207/10012 Stereo images
• CPC Class Codes: G06T 2207/30242 Counting objects in image
• CPC Class Codes: G06T 7/85 Stereo camera calibration
• CPC Class Codes: H04N 13/246 Calibration of cameras

Abstract

In a method for calibration of a stereo camera (1) comprising two cameras (2, 3) arranged in a distance along a first direction (x) and mechanically attached to a common structure, rectification functions mapping raw image coordinates to rectified coordinates are provided. Recalibration is based on images of a scene captured by the stereo camera (1) affixed in an observation position, the recalibration yielding correction functions mapping rectified coordinates to corrected coordinates. Recalibration is based on matching objects comprised in images of the captured scene obtained by the two cameras (2, 3). The stereo camera (1) is affixed in a fixed relationship with a planar surface (4), with a known distance and angle between the stereo camera (1) and the surface (4), and objects are matched in the recalibration that belong to the surface (4), wherein a disparity caused by the known distance is exploited when determining the correction functions. Exploiting the additional information with respect to the surface (4), the determination of the correction functions may be greatly improved. In particular, the inventive method allows for employing general correction functions, and no modelling of the calibration errors is required.

Claims

1. A method for calibration of a stereo camera comprising two cameras arranged in a distance along a first direction and mechanically attached to a common structure, the method comprising the steps of:

a) providing rectification functions mapping raw image coordinates to rectified coordinates;

b) recalibration based on images of a scene captured by the stereo camera affixed in an observation position, the recalibration yielding correction functions mapping rectified coordinates to corrected coordinates;

where

c) recalibration is based on matching objects comprised in images of the captured scene obtained by the two cameras;

wherein

d) the stereo camera is affixed in a fixed relationship with a planar surface, with a known distance and angle between the stereo camera and the surface, and in that

e) in the recalibration objects are matched that belong to the surface, wherein a disparity caused by the known distance is exploited when determining the correction functions.

2. The method as recited in claim 1, wherein a first correction function for obtaining a corrected coordinate along the first direction and a second correction function for obtaining a corrected coordinate along a second direction perpendicular to the first direction both depend from a first coordinate along the first direction and a second coordinate along the second direction.

3. The method as recited in claim 2, wherein the first correction function and the second correction function are second order polynomials of the first and the second coordinate.

4. The method as recited in claim 1, wherein the recalibration is based on rectified coordinates.

5. The method as recited in claim 1, wherein a field of view of the stereo camera is at least 90°

×

80°

.

6. The method as recited in claim 1, wherein a lens distortion in the used lens area is at least 0.2%.

7. The method as recited in claim 1, wherein the planar surface is a floor surface and in that the observation position is above the floor surface.

8. The method as recited in claim 7, wherein the distance between the stereo camera and the surface amounts to at least 2.20 m.

9. The method as recited in claim 1, wherein an angle between the stereo camera and the surface is measured by a sensor device integrated to the stereo camera.

10. The method as recited in claim 1, wherein the objects comprised in images of the captured scene are matched by a feature extraction algorithm.

11. The method as recited in claim 1, wherein matching of the objects comprised in images of the captured scene is done with subpixel precision.

12. The method as recited in claim 1, wherein the correction functions are obtained by regression analysis, optimizing an agreement between coordinates of the matched objects in an image obtained by a first of the two cameras with coordinates of the matched objects obtained from applying the correction functions to the coordinates of the matched objects in an image obtained by a second of the two cameras.

13. The method as recited in claim 12, wherein the coordinates of the matched objects in the image obtained by the first camera are shifted by the expected disparity due to the known distance.