Automatic registration of multi-source medium resolution satellite data

Abstract

Multi-temporal and multi-source images gathered from satellite platforms are nowadays a fundamental source of information in several domains. One of the main challenges in the fusion of different data sets consists in the registration issue, i.e., the integration into the same framework of images collected with different spatial resolution and acquisition geometry. This paper presents a novel methodology to accomplish this task on the basis of a method that stands out from existing approaches. The whole data (time series) set is simultaneously co-registered with a two-dimensional multiple Least Squares adjustment with different geometric transformations implemented. Some tests were carried out with different geometric transformation models (including similarity, affine, and polynomial) and variable matching thresholds. They showed a sub-pixel precision after the computation of multiple adjustment. The use of multi-image corresponding points allowed the improvement of the registration accuracy and reliability of a time series made up of data imaged with different sensors.