Depth camera calibration using depth measurements

Pahwa, Ramanpreet

Permalink

https://hdl.handle.net/2142/45544 Copy

Description

Title

Depth camera calibration using depth measurements

Author(s)

Pahwa, Ramanpreet

Issue Date

2013-08-22T16:47:15Z

Director of Research (if dissertation) or Advisor (if thesis)

Do, Minh N.

Department of Study

Electrical & Computer Eng

Discipline

Electrical & Computer Engr

Degree Granting Institution

University of Illinois at Urbana-Champaign

Degree Name

M.S.

Degree Level

Thesis

Keyword(s)

• Depth Cameras
• Calibration. Quaternions
• cross-calibration
• self-calibration

Abstract

An important recent development in the visual information acquisition field is the emergence of low cost depth cameras that measure the scalar distance between the objects present in the scene and the cameras. These cameras project infra-red rays and use time of flight to measure the distance at each pixel. These cameras have the potential to significantly impact various computer vision fields. However, due to the presence of significant noise and low resolution of such cameras, their impact is currently extremely limited. These cameras need to be calibrated accurately before they can be used along with color cameras to perform various tasks such as 3D reconstruction and augmented virtual reality. In this thesis, we propose to use the measurements provided by the depth cameras - depth and estimated intensity at each pixel to de-noise the depth images - and then use them for calibration. Previous work to calibrate the depth cameras involves either using a multiple camera set-up or using an extremely precise optical measurement rack to move checker-board images in the camera’s field of view. However, we want average users to be able to calibrate these cameras at home without having access to such precise instruments. Thus, we are motivated to consider an efficient and robust calibration scheme that only uses the measurements taken from the depth camera and a simple checker-board that a user waves in front of the camera. This thesis characterizes the noise present in depth measurements that are captured by the ToF cameras. We propose a thresholding and a grid based planarization scheme to de-noise the depth images before we use these measurements for camera self-calibration. We utilize a two-step non-linear optimization technique utilizing the Levenberg-Marquardt Algorithm (LMA) to minimize the projected distance between measured and computed corner points in each checker-board image. We also propose a new method using quaternions for automated cross-calibration between a depth and color camera that calibrates both devices and estimates the intrinsic and extrinsic parameters of the cameras without using any known geometry in the scene. Our results demonstrate that the quaternion approach provides results as good as those of existing techniques but 3-5 times faster.

Graduation Semester

2013-08

Permalink

http://hdl.handle.net/2142/45544

Copyright and License Information

Copyright 2013 Ramanpreet Pahwa

Owning Collections