Roshan Suresh Kumar, FNU (2019-04). Calibration of Linear Imager Camera for Relative Pose Estimation. Master's Thesis. Thesis uri icon

abstract

  • The process of camera calibration is of paramount importance in order to employ any vision based sensor for relative navigation purposes. Understanding and quantifying the physical process that converts the external electromagnetic stimulus into an image inside a camera is key to relating the position of a body in an image to its pose in the real world. Both camera calibration and relative navigation are extensively explored topics. In the topic of camera calibration, various algorithms have been proposed that model the image formation process in different ways. This research utilizes the Homography approach proposed by Zhang [1] along with two distortion models: Brown's nonlinear Distortion Model and the Geometric Distortion Model in order to model the intrinsic distortion and discrete image formation process. The idea of this research is to utilize the intrinsic parameters estimated using the homography optimization approach for the estimation of the relative pose of an object in the camera's field of view. A nonlinear optimization based approach is presented for this purpose. The camera used here is the Phasespace Motion Capture camera [2] which utilizes linear imagers to form a fictitious image plane. Hence, the applicability of the two distortion models is tested through multiple datasets. Through testing with three datasets, it is found that neither distortion model is adequate to describe the distortion and image formation process in the Phasespace camera. A further test is conducted in order to validate the efficacy of the optimization based approach for relative pose estimation.

publication date

  • May 2019