Can a model be accurately fitted to data with a high percentage of errors? This paper introduces Random Sample Consensus (RANSAC), a new paradigm designed for interpreting and smoothing experimental data containing significant gross errors. RANSAC is well-suited for automated image analysis, where interpretation relies on data from error-prone feature detectors. A major portion of the paper focuses on applying RANSAC to the Location Determination Problem (LDP): determining the point in space from which an image was obtained, given known landmark locations. In response to RANSAC requirements, the paper derives new results on the minimum number of landmarks needed and presents algorithms for computing these solutions in closed form. These results provide the foundation for an automatic system capable of solving the LDP under challenging viewing conditions, with implications for computer vision and robotics.
Published in Communications of the ACM, this paper presenting the RANSAC algorithm aligns with the journal's focus on fundamental algorithms and their applications in computer science. The work has had a lasting impact on the field of computer vision and related areas.