Zusammenfassung
Summary
In recent years, there has been a strong upswing in the construction industry especially with respect to residential buildings. Due to the obligation to document changes in the building stock in Bavaria, this article presents a new method for the detection of construction activities from remote sensing data. The approach bases on time series from freely available multi-temporal active Sentinel-1 (S1) and passive Sentinel-2 (S2) satellite images from the European Copernicus programme. Data collected during the period between the beginning of October 2019 and the end of November 2020 over a study area in the town of Dingolfing (Lower Bavaria, Germany) were evaluated. After processing the multi-temporal polarimetric and optical images to Analysis Ready Data (ARD) in the Kennaugh framework and applying the novel LEAIndex (LEAI), a simple, but robust, Wavelet-based approach for the detection of new buildings was developed. The validation of the results towards in-situ owner information gathered by the help of a questionnaire revealed that out of 38 new buildings 34 were reliably identified by the prototypic algorithm. The described methodology is intended to be implemented in a steady detection service for construction activities.