Detection of small changes in airborne hyperspectral imagery: Experimental results over urban areas

Abstract

In this work we investigate the problem of detecting small changes in images acquired by airborne sensors, using direct georeferencing from gyro data and GPS position. We intend to avoid the time consuming step of image registration, exploiting direct georeferencing cascaded with a robust change detection strategy that can properly manage the typical registration errors given by onboard instrumentation. We investigate the effectiveness of this approach in the urban scenarios, where we are interested in detecting changes induced by small objects. The experimental analysis conducted on real hyperspectral data with very high spatial resolution highlights the effectiveness of the proposed approach, resulting in a consistent improvement of both the capability of detecting changes and of suppressing the background.