This work has been realized with the belief function theory (or Dempster-Shafer theory), a model for uncertainty representation which generalizes in particular the probability theory. In this model, a summary of our work related with correction mechanisms of the information provided by a source is given, as well as an application in classification. Our contributions in three applications in information fusion with belief functions are also exposed. These applications concern car to car communication, data association with belief functions, the diagnosis of the winding ageing of an electric machine.