On depth usage for a lightened visual SLAM in small environments
Abstract
Historically popular, the well established monocular-SLAM is however subject to some limitations. The advent of cheap depth sensors allowed to circumvent some of these. Related methods frequently focus heavily on depth data. However these sensors have their own weaknesses. In some cases it is more appropriate to use both intensity and depth informations equally. We first conduct a few experiments in optimal conditions to determine how to use good quality information in our monocular based SLAM. From this we propose a lightweight SLAM designed for small constrained environments.