On-the-fly Lens Calibration for Image Navigation and Map-aided Navigation

Project Title: On-the-fly Lens Calibration for Image Navigation and Map-aided Navigation

Project Description: Synchronized stereo video-based image dead reckoning navigation systems experience performance degradation due to the two lenses becoming uncalibrated (unrectified). This causes the two image planes of the two cameras misalign from the same plane such that optical axes of the two image planes are no longer parallel. The resulting left and right image geometric relationship can no longer be used to compute the 3-D position of features accurately.

Our team has designed a realtime auto-calibration procedure to restore proper stereo camera calibration for synchronized stereo video-based image dead reckoning navigation systems. We are now extending this to monocular-based image dead reckoning systems – in which there is only one image sensor so there is no need to align two image planes, however, rectifying is still necessary to be able to accurately relate image features and real world geometry. It is desirable to have realtime lens calibration techniques for both stereo-based and monocular image dead reckoning systems. Future applications are envisioned where an image sensor, not originally designed or deployed for navigation use, is utilized for this purpose where calibration will need to be made without any prior calibration, knowledge or model of the lens.

Anticipated Students Tasks and Responsibilities:

  1. Understand camera calibration.
  2. Investigate factors that lead to camera miscalibration and characterize the effects on all the relevant camera parameters.
  3. Help develop and design algorithms for monocular on-the-fly lens calibration, including correcting for radial distortion.

Qualifications: The candidate should have experience working with digital video-cameras (preferably industrial ones) and understand the underlying concepts such as projective camera geometry, rectification, calibration, image formation and acquisition. Background in image processing/computer vision is preferable, which implies understanding of linear algebra concepts. Good programming skills are required (MATLAB and C/C++).

Appropriate Student Majors or Programs: Computer Engineering (CprE), Electrical Engineering (EE)

Undergraduate Research Assistant Position For: Fall and Spring

Hours of Work Offered Per Week: 10

Paid or Unpaid: Paid

Professor: Arun K. Somani

How to Apply: E-mail resume and cover letter describing why you are suited for this project.