A Feature-Based Scene Matcher

A method is presented for matching two scene descr ip t ions, each of which consists of a set of measured feature vectors w i th estimated uncer ta int i e s . The two scenes d i f f e r by a transformation that depends on a few unknown parameters. The method performs a search by sequent ia l ly matching features of one scene to those of the other scene, solv ing for the transformation parameters by means of a generalized least-squares adjustment, computing the p robab i l i t i e s of these matches by means of Bayes* theorem, and using these p robab i l i t i es to prune the search. An example is given using scene descr ipt ions of the Martian surface in which the features are rocks approximated by e l l i p so i ds .