Graphics and Media Lab

Downloads

v.0.1 18.06.2008 Download GML AeroMatching 0.1

Introduction

GML AeroMatching is a free tool for aerial images matching. It provides a rotation-invariant algorithm, flexible enough to be applied to various types of aerial images. Such settings as initial number of points used for further matching and descriptor patch radius are available.

The algorithm uses no information about the camera and doesn't need any initial approximation. Recommended image overlap area is more than 50%.

Test data has been provided by GeoLidar company. The photos were taken with the "Rollei AIC modular 5*7" aerocamera.

The code was written with the use of OpenCV 1.0 and was tested mostly on win32 (Win2K, WinXP) platform.

Algorithm

Aerial images matching algorithm includes the following steps:

• Creation of an image scale pyramid;
• Feature points extraction using Harris corner detector [1] on the roughest image pyramid layer;
• Computation of invariant features for each extracted point [2];
• Pairwise invariant features matching;
• Robust homography matrix estimation using RANSAC techniques;
• Rejection of false matches by considering the homography matrix.

Requirements

System Requirements: .Net Framework 2.0
At least 1024x768 screen resolution is recommended.

References

[1] Harris C. and M.J. Stephens. A combined corner and edge detector. In AlveyVision Conference, pages 147–152, 1988.

[2] J. Flusser, "On the Independence of Rotation Moment Invariants," Pattern Recognition, vol. 33, pp. 1405-1410, 2000.

[3] P.Torr, A.Zisserman, “Robust Computation and Parametrization of Multiple View Relations”, ICCV Proc, p.727, 1998 

Team

• Anton S. Konushin, Lomonosov Moscow State University
• Alexander Velizhev, Lomonosov Moscow State University
• Gleb Krivovyaz, Lomonosov Moscow State University
• Alexey Chernyavskiy