Detectors and Descriptors for Photogrammetric Applications

Abstract

This paper reports about interest operators, region detectors and region descriptors for photogrammetric applications. Features are the primary input for many applications like registration, 3D reconstruction, motion tracking, robot navigation, etc. Nowadays many detectors and descriptors algorithms are available, providing corners, edges and regions of interest together with n-dimensional vectors useful in matching procedures. The main algorithms are here described and analyzed, together with their proprieties. Experiments concerning the repeatability, localization accuracy and quantitative analysis are performed and reported. Details on how improve to location accuracy of region detectors are also reported.