Service Communautaire d'Information sur la Recherche et le Développement - CORDIS

System for wide baseline matching of images from autonomous aerial vehicles using blob features

The system can match images from widely separated views of the scene, possibly from different UAV’s. The implementation is based on robust blob features. Blobs are perceptually salient, homogeneous, and compact image regions. Each blob is characterized by its colour, its center-point and its shape represented by the inertia matrix.

First, correspondences between blobs are obtained by using the mean colour. After that, a RANSAC algorithm refines these correspondences using the position of the blobs. Also, a shape metric derived from the inertia matrix of each blob is used. As a result, a set of correspondences are found and a homography model relating the two views is obtained.

Reported by

Linkoepings Universitet
Computer Vision Laboratory, Department of electrical engineering, Linköping University
581 83 Linkoeping