Abstract

The analysis of land cover change plays a key role in understanding a great variety of phenomena in several research fields and aerial photos represent the main existent database providing evidence of landscape changes with high detail. However, the extraction of information from aerial photos can be really harsh especially when dealing with historical photos and/or when involving areas with a complex topography. This work presents a semi-automated object-oriented procedure for the analysis of landscape dynamics through the comparison of land cover maps derived from actual and historical aerial photos. The same minimum map unit (40 m²), the same thematic detail and the geometric correspondence between stable landscape features (roads and buildings) allowed a good comparability of the resultant maps (1954 and 1999). Moreover, the quality of some important steps, as radiometric correction and segmentation set-up, has been evaluated quantitatively. Finally, the land cover change analysis confirmed the presence of characteristic processes of the Mediterranean mountains, as forest recovery and landscape homogenization. In conclusion, the developed approach satisfies the requests of objectivity, replicability, cost minimization and large applicability and it deepens the comprehension of the multi-faceted map building processes providing a useful tool to improve and delve into territorial studies.

Key words: GAI flight, orthorectification, radiometric correction, object-oriented segmentation,maps building, land abandonment.