Structuring the Components of the Image Matching Problem
This Note briefly describes the matching process. The major emphasis is in describing the elements of the matching process — the scene, matching algorithms, and errors — and determining their roles in and effect on the matching process. A means is provided for structuring the map matching problem. The scene is defined by the degree of homogeneity and the number of independent elements in each homogeneous region. The errors are further broken up into categories that are mutually exclusive, comprehensive, and positively related to a preprocessing technique or algorithm required to accommodate them. The errors are thus broken up into one of the following categories: global, regional, local, and nonstructured. Finally, the matching algorithms are defined as being of a feature matching correlation or hybrid type. The latter type is a new class of algorithm developed at RAND which bridges the gap between feature matching and correlation types of algorithms.