Improving visual SLAM algorithms for use in realtime robotic applications

Many vision-based Simultaneous Localization And Mapping (vSLAM) algorithms require large amounts of computational power and storage. With these requirements, vSLAM is difficult to implement in real time. One known bottleneck in vSLAM is performing feature identification and matching across a large database. In this paper, we present a system and algorithms to reduce computational time and storage requirements for feature identification and matching components of vSLAM. We compare our algorithms using ORB and SURF to their unmodified versions readily available datasets and show significant reductions in storage requirements and calculation time.