부산대학교 도서관

This paper describes the Automated Registration Evaluation System (ARES). The goal of the ARES is to dramatically reduce the cost of evaluating different automated data (raster and/or vector) registration technologies. Under the first year of the ARES project,five different automated registration methods were selected for evaluation. These methods represented the state-of-the-art in automated image-to-image registration and image-to-3D feature registration. During the first year of testing, over two terabytes of commercial imagery and GIS feature (vector) data were acquired. ARES performed over 4000 individual registration test cases among the five different automated registration methods. ARES was able to ascertain the operating conditions and performance of each of these methods against a wide range of user requirements. Results from ARES suggest that automated image-to-image matching for near-nadir satellite imagery (with good initial approximations) is essentially a solved research problem, whereas feature-to-image matching is not.