A Semi-Automated Framework for Homography Estimation

Thomas Fäulhammer and Paulo V.K. Borges

Keywords

Camera Calibration, Homography Estimation

Abstract

Reliable estimation of the geometric transformation (homography) between different planes is a key step in a number of robotics applications. A common requirement is the homography estimation between a ground plane and a camera. For this task, the most usual approach is to define the correspondence between the coordinates of real-world points and their pixel coordinates in the image. Often, however, real-world coordinates are not available or ground measurements are not practical. Targeting these scenarios, a homography estimation framework is proposed using (semi-)rigid objects that can be placed in different positions in the field of view, with its top and bottom pixels approximately identified in the image. The derived method shows that it can work with or without the knowledge of the camera intrinsics, even in presence of non-Gaussian measurement noise. A thorough analysis of the method is presented by observing the effect of noise, distance from the camera and number of measurements on the quality of the final results. The experiments illustrate the applicability of the method.

Important Links:

Go Back