Random Sample Consensus (RANSAC) is a new paradigm for fitting a model to experimental data, particularly useful for handling data with a significant percentage of gross errors. The method is especially applicable to image analysis and automated cartography, where data from error-prone feature detectors may contain outliers. The paper introduces RANSAC and applies it to the Location Determination Problem (LDP), which involves determining the spatial location from which an image of known landmarks was taken. RANSAC is capable of identifying a subset of data points that are consistent with a model, allowing for robust model fitting even in the presence of outliers. The LDP is formally defined as determining the location of the camera (Center of Perspective) relative to known landmarks in an image. The paper presents new results on the minimum number of landmarks required to solve the LDP and describes algorithms for computing these solutions in closed form. These results form the basis for an automatic system that can solve the LDP under difficult viewing and analysis conditions, even when a significant number of landmarks are incorrectly located. RANSAC works by randomly selecting subsets of data points to instantiate a model, then identifying a subset of points that are consistent with the model. If a sufficient number of consistent points are found, the model is refined using techniques like least squares. The process is repeated until a consensus set is found or a predetermined number of trials is reached. The paper also discusses the theoretical foundations of RANSAC, including error tolerance, the number of trials required to find a consensus set, and the size of an acceptable consensus set. The paper presents an example of RANSAC applied to the LDP, demonstrating its effectiveness in handling gross errors. It also describes the solution to the Perspective-n-Point (PnP) problem, which involves determining the location of the camera given the relative positions of known landmarks in an image. The paper shows that the P3P problem (three points) can have up to four solutions, while the P4P problem (four points) can have multiple solutions depending on the configuration of the points. The paper concludes that RANSAC provides a robust method for solving the LDP and other similar problems in image analysis and automated cartography.Random Sample Consensus (RANSAC) is a new paradigm for fitting a model to experimental data, particularly useful for handling data with a significant percentage of gross errors. The method is especially applicable to image analysis and automated cartography, where data from error-prone feature detectors may contain outliers. The paper introduces RANSAC and applies it to the Location Determination Problem (LDP), which involves determining the spatial location from which an image of known landmarks was taken. RANSAC is capable of identifying a subset of data points that are consistent with a model, allowing for robust model fitting even in the presence of outliers. The LDP is formally defined as determining the location of the camera (Center of Perspective) relative to known landmarks in an image. The paper presents new results on the minimum number of landmarks required to solve the LDP and describes algorithms for computing these solutions in closed form. These results form the basis for an automatic system that can solve the LDP under difficult viewing and analysis conditions, even when a significant number of landmarks are incorrectly located. RANSAC works by randomly selecting subsets of data points to instantiate a model, then identifying a subset of points that are consistent with the model. If a sufficient number of consistent points are found, the model is refined using techniques like least squares. The process is repeated until a consensus set is found or a predetermined number of trials is reached. The paper also discusses the theoretical foundations of RANSAC, including error tolerance, the number of trials required to find a consensus set, and the size of an acceptable consensus set. The paper presents an example of RANSAC applied to the LDP, demonstrating its effectiveness in handling gross errors. It also describes the solution to the Perspective-n-Point (PnP) problem, which involves determining the location of the camera given the relative positions of known landmarks in an image. The paper shows that the P3P problem (three points) can have up to four solutions, while the P4P problem (four points) can have multiple solutions depending on the configuration of the points. The paper concludes that RANSAC provides a robust method for solving the LDP and other similar problems in image analysis and automated cartography.