Models and their matching contexts (focus features and associated groups) can either be read from files (see Appendix for a definition of the format of these files), generated at run time, or a combination of both (the model read from disc and the allowable matching contexts determined at run time). Once the model has been obtained the process of matching it against a current scene description (assuming it has already been obtained in the usual way using the Edge tool) has 2 stages. The first is to build a pairwise relations table for the model and scene; this contains information about the invariant relationships between pairs of features. The second is to perform a constrained search over the tables of the model and scene to identify cliques of above threshold cardinality between matching features. These are used to compute potential transforms and recruit further matches. The model matcher has specialised File/View facilities for reading/writing, and display/manipulation of scene and model data. Using the view facility and associated Tv devices (model andscene) the matched model can be displayed in the transformed position over the scene description. The matcher has 2 parameter cycles and toggle selectors for adjusting various pair wise relationships table and matching algorithm parameters.