Vol. 11, No. 1, 44-51, 2012

Abstract
This paper compares a pixel-based vs. an object-based classification (OBIA) of a multi-spectral IKONOS image for mapping Mediterranean forests after fire. The study area was a part of the aesthetic, peri-urban forest of Thessaloniki, Greece (Seich-Sou forest) after the big fire of 1997. The IKONOS image was acquired in 2001 and was classified with Maximum Likelihood Classification (MLC, per-pixel method) and with the Membership Function classifier (per-object method). The latter relies on the development of rules, which have the potential to support objective and standardised mapping. Five land use/cover classes were recognised in the forest according to the national legislation. Verified by visual photointerpretation and ancillary ground truth data, OBIA seems to give more realistic results than MLC, especially for grasslands, agricultural land and reforested areas. Future work will focus on testing OBIA in the entire extent of the forest towards an operational use of the rule-set for monitoring Mediterranean forests after fire.