An Object-Oriented Spatial Information Extraction from IKONOS Multispectral Data to Detect Agricultural Transformations

Desertification is increasingly becoming a problem within the Mediterranean Basin Countries, threatening 27% of the Italian territory. In the Apulia region, and particularly in the Alta Murgia area, rock-breaking practices must be added to the list of land degradation causes. The custom of removing stones has been practised for a long time and consists in removing outcrops of stone from cultivated fields at the end of every productive cycle, for the development of rural constructions. In recent years such custom has been substituted by rock-breaking practices, that consist in the transformation of pasture areas into cultivated land by means of deep ploughing and mechanical crushing of rocks. In this work an object-oriented image analysis (OBIA) on IKONOS multispectral data was implemented for a sub-set area of the Alta Murgia National Park. The analysis carried out consisted of two phases: fist, a segmentation of the image into meaningful multipixel objects of various sizes, based on both spectral and spatial characteristics of groups of pixels; at a second stage, the segments (objects) were assigned to classes using fuzzy logic and a hierarchical decision key. A comparison between the results from the ISODATA classification and the OBIA was also given.