Given the large ecological changes of the last decade and the importance of carbon stocks underlined by the Kyoto Protocol, change detection in forest surfaces has become very essential. Thanks to a growing multiplicity and accessibility of satellite data, new methods are required to exploit these data at best and then provide a continuous monitoring of the environment. However, there is now no robust method to detect changes in forest areas. The aim of this research is the development of an automatic method to detect and interpret land cover changes in forested areas using high resolution satellite images. The double originality of this study is (i) the object-based approach used instead of classical pixel-based methods, and (ii) the automation of the process. This technique will be tested in different forest environment, both temperate and tropical, in order to generalize the process.