A VSLAM Fusing Visible Images and Infrared Images from RGB-D camera for Indoor Mobile Robot

Abstract

Vision-based simultaneous localization and mapping (VSLAM) system cannot work well in low illumination or darkness, because there is not enough information in a visible image taken in that situation. In this paper, a method is presented to solve this problem by fusing information of visible images and infrared images. The key idea is to create map points respectively from visible images and infrared images for location and mapping. Evaluation in 4 sequences in 3 illumination conditions shows that our method is performed well.