Evaluation of feature detectors for registering aerial images
Aaron Marburg, Michael Hayes, Andrew Bainbridge-Smith · 2012
The detection, extraction, and matching of image features is a popular method for generating point-to-point correspondences for the estimation of scene and camera geometries. In this work we evaluate the performance of a variety of feature detection algorithms over two reference data sets and a set of aerial images which includes large changes in scene illumination. The evaluated detectors showed expected performance against the reference data sets, and aerial images with constant lighting conditions, but were unsuccessful in aligning image pairs showing strong changes in image exposure and illumination.