Research on object recognition invariant under out-of-plane rotations has so far yielded limited results. The problem becomes even more complex when in addition scale changes must also be taken into account. We develop a new object recognition method invariant to translations, rotations, changes of pose, and scale. The method is based on angular wedge sampling about the centroid of the object, yielding translation-, rotation-, and scale-invariant features. A modified feature space trajectory classifier is used to obtain out-of-plane rotation invariance. The method is successfully tested on models of military land vehicles and is optically implementable.