Abstract

Remote sensing techniques provide images of large areas of the earth’s surface at relatively frequent intervals. During the last decade, technological progress has substantially influenced and improved the availability and analysis of satellite and other earth observation data on the one hand; on the other hand, the demand for up-to-date data by diverse users has increased. Therefore, earth observation has become an indispensable tool in various fields and is also coming into play also in official statistics.

To support the construction statistics by quality assurance measures, the project ‘Earth Observation and AI for Construction Statistics’ (EO4ConStat) was set up of a consortium of the German National Statistical Institute (Destatis), the Federal Agency for Cartography and Geodesy (BKG) and the German Aerospace Centre (DLR). The objective of developing a method to quality assure to construction activity statistics using earth observation data and artificial intelligence is relevant for high quality construction statistics.

Due to the German government's ambition to build new, affordable and climate appropriate housing, construction statistics are currently in the political focus. The aim of the project – which is financed by the European Commission – includes developing algorithms to detect buildings and building construction sites and possibly define new construction starts as well as works completions from earth observation data. This is to be achieved by using and adapting a foundation model. The used model is a permissive open license segmentation algorithm that is finetuned to detecting construction sites and stages of construction works. Moreover, with a time series analysis some breaking points shall be detected in the considered data. Finally, a methodology to compare the results of the segmentation algorithm and the change detection analysis with the data collected through the traditional statistical channels shall be contrived.