Geoinformatics is becoming increasingly important in all areas of life, helping to
speed up many time-consuming tasks. In our research, we compared a WorldView-2 (WV2),
WorldView-3 (WV3) satellite image and a hyperspectral aerial image for roof classification
in the case of 4 categories: asbestos, red tiles, red bitumen and asbestos covered
with red bitumen. We used Random Forest (RF) and Support Vector Machine (SVM) classifiers
in Python. We also performed a change assessment for asbestos roofs between 2013 and
2019. The best results were obtained for asbestos (F1 scores above 93%). There were
significant misclassifications for bitumen roofs. Although the number of asbestos
roofs decreased over the period, in 40% of cases these roofs were simply covered with
bitumen shingles rather than being replaced.
