The application of Random Sample Consensus in photogrammetric relative orientation

This paper proposes a method based on the rule of Random Sample Consensus (RANSAC) to locate and extract gross errors in relative orientation. This method uses the coplanarity condition of stereo image pair as model to fit the data. With this method we do several experiments with different size gross errors to validate the effectiveness of it. The experiment results show the mismatching points with different size are eliminated. This improves the internal reliability of relative orientation.