An improved Random Sample Consensus algorithm for visual indoor positioning based on angle difference

The Random Sample Consensus (RANSAC) algorithm is a classical parameter estimation and model fitting algorithm, which suffers from poor accuracy and robustness in removing mismatches when dealing with images in complex indoor environments. In this paper, an improvement method of RANSAC is proposed: the mismatch probability of each pair of feature points is quantified by analyzing the angular difference between the SURF feature descriptors, and the mismatch probability is weighted to the stochastic process of RANSAC, in order to improve the accuracy of the matching and the robustness of the algorithm. The experimental results show that the improved RANSAC algorithm is more effective in attenuating the mismatch of images; from the positioning results, the average positioning error obtained by the traditional RANSAC algorithm is 1.09 m, and the average positioning error obtained by using the improved RANSAC algorithm is 0.66 m. Moreover, the probability of the improved method in the positioning experiments with an error of less than 1 m reaches 89%, which effectively improves the positioning accuracy and robustness.