Projects

Bayes

Delaunay

Finger

Homography

Membrane

Multi Cam

Pong

Pool

Robot

Skull

Stereo

Wing

Homography

(using C, MFC, DirectShow and OpenCV)
Return
The aim of this project is to estimate the 3D trajectory of an aircraft from 2D pictures (from camera on it).

  1. Introduction

    This training work is related to the project .Vision for Aircraft Landing.. More exactly, the aim is guiding in an automatic way an aircraft provided with a camera using dynamic vision (without any real time restriction at this stage).

    The position and orientation of the aircraft can be partially estimated using visual informations acquired by a camera mounted on the aircraft.

  2. Homography estimation

    • Least squares estimation of the homography
    • Singular values decomposition (SVD)
      In each of the two solutions, sgn(.) = - sgn(.), and w is an arbitrary coefficient (we will set it at 1). Let us notice that only the orientation of the planar patch is estimated. Its position is not known.

    • To solve the ambiguity problem, i.e. to choose the right solution from above, Faugeras and Lustman suggest three ways to proceed [Faugeras & Lustman 88]:
      1. Look at a second plane.
      2. Use a third image.
      3. Use relationships between features in the plane.

  3. Pose-estimation algorithm

  4. Aircraft pose-estimation interface

  5. Homography estimation

    Images from the aircraft Estimated trajectory


    This animation is just to be able to understand the results (the animation is not right).