Vision Feature Extraction Algorithm for Occupancy Grid Maps Merging

Abstract

With the improvement of single robot SLAM technology, multi-robot SLAM became a research hotspot.Map merge is one of the difficult issue.In this paper, aORBfeature extraction algorithmfor occupancygrid maps mergingwasimproved. ORB features have good rotation and scaleinvariance, while maintaining a compromise calculation speed, in the real-time SLAMaccess to a large number ofapplications.Based on Robot Operating System,two omnidirectional robots equipped with a 2D laser sensor RPLIDARbuildgrid maps via Hector SLAM. Map mergeisintegrated as an image registration problem, by extracting the ORB features of the grid mapsand usingBruteForce search method to match, among thebest matches,calculatingaffine transformationand merging.The experimental results show that the improved algorithm has abetter performance on real-time and effectiveness.