LAUSR

ABSTRACT In this article, we propose a methodology to co-register multi-temporal images captured from different sources in order to make it possible to generate precise, fully-automatic spatial information. The goal is to identify objects in images without metric or assigned coordinates by relating them with the same objects in images from other sources where the metric is known. In this way, processing times are reduced and manual intervention is unnecessary, thus making it ideally suited for continuous update programmes. This article describes the use of a modified optimization of the scale invariant feature transform to successfully match old and new images. After that, control points are automatically assigned to each new image to generate orthophotos automatically and evaluate them through positional accuracy tests.