Hi all, i have a satellite image of a city and i want to classify every object in the image as vegetation, buildings, roads, rivers, etc . My idea is to do segmentation upon the image then classify it using bayes classifier, could anyone give me a few hints ?

A good approach
Sounds like a good approach. If you have ground truth (eg segmented and hand-labeled), it should be straightforward to build a simple Bayes classifier based on local texture. In fact you could probably skip the segmentation step at classification-time and just classify each local neighbourhood.