This paper describes a feature point matching strategy and motion recovery applied to vehicle navigation. A transformation of the image plane is used that keeps the motion of the vehicle on a plane parallel to the transformed plane. This allows us to define linear tracking filters to estimate the real-world positions of the features. The correspondences between features are first selected by similarity, taking into account the smoothness of motion and rigidity of the scene. Further processing brings out the rest of correspondences. The method is applied to a real application consisting of an autonomous vehicle navigating in a crop field.