PRINCIPLES AND APPLICATIONS OF STRUCTURAL IMAGE MATCHING

Technological developments in stereo image matching have advanced from traditional area-based to feature-based matching and lately to struct ural matching. This paper introduces a newly developed matching method utilizing structural matching. An evaluation function based on the jo int probability of a possible match is derived. An informed tree searc h method is applied to find the best match. In order to keep the searc h time acceptable for photogrammetric applications, a series of strate gies have been developed and integrated into the search method. Proced ures for extraction of structural descriptions and an algorithm for ro tation-invariant correlation are presented. In combining all these too ls, the fully automated matching of image features is realized without any a priori information. Several tasks like relative image orientati on, image matching of non-metric images and tie point measurement in a erotriangulation, which were fully automated by using the structural m atching approach, are presented. The results indicate that structural matching techniques can achieve a very high level of automation in dig ital photogrammetry. (C) 1998 Elsevier Science B.V. All rights reserve d.