Scalable and Error Tolerant Automated Georeferencing under Affine Transformations

Georeferencing is a key component in GIS systems. However, its performance has been dragged by the manual georeferencing process. In this paper, we develop a scalable and error tolerant automated georeferencing scheme for general affine transformations called localized matching and global optimization (LM-GO). Under LM-GO, the complexity of control point pair (CPP) matching is not directly tied to image or vector data size, which makes it highly scalable. The unavoidable image point extraction errors are effectively managed and filtered out by the verification and optimization process. Another major contribution of this paper is to propose the first asymmetric CPP matching algorithm for general affine transformations, namely AIPP matching, based on pattern matching methodology, which improves CPP matching performance by an order. Overall, the running time of LM-GO grows almost linearly with vector data size. LM-GO effectively reduces the georeferencing time from hours by the bruteforce search approach down to seconds.