An Application of Improved RANSAC Algorithm in Visual Positioning

Abstract

In the visual positioning of mobile robots, image feature matching technology is constantly developing, but there are still problems such as low real-time performance and low precision. Therefore, the feature matching is improved by the combination of the traditional ORB algorithm and the RANSAC (Random Sampling Consensus) algorithm. On the basis of the Hamming distance, the improved RANSAC algorithm pre-processes the mismatched point pairs according to the characteristics of the distance, rotation and angle consistency of the correct matching point pairs. Secondly, the Gaussian function is used as the threshold value, and the data is classified according to the weights. The quadratic optimization prevents the RANSAC algorithm from falling into local optimum, which can achieve image matching fastly and accurately. The simulation results of visual positioning indicate that the proposed algorithm can improve the accuracy of image matching and reduce the time of image matching effectively. The estimated trajectory of this algorithm is smoother and more fitting than traditional algorithms.