A mobile robot repositioning method combining vision and laser

Abstract

The laser sensor information is single, the perceptual feedback dimension is low, and the feedback information is lack in the open outdoor scene and the scene with similar geometric features, which makes it difficult to complete the positioning and navigation work. The visual positioning is easily interfered by light, and the accuracy is limited by the algorithm. Combining the advantages and disadvantages of the two methods, an algorithm combining vision and laser repositioning is proposed. Firstly, vision is based on ORB feature matching to do the preliminary localization, and an artificial bird swarm algorithm is proposed to improve the accuracy of local localization when the prior location is known. The experimental results show that the proposed algorithm shortens the relocation time by one order of magnitude compared with the pure laser localization algorithm, and improves the localization accuracy by one order of magnitude compared with the pure vision algorithm.