A Tracking and Mapping Method for Visually-degraded Environment

Abstract

When exploring uncertain areas, the system usually relies on Simultaneous Localization and Mapping (SLAM) to map the surrounding environment and track the location through the environment. Recent work has shown that the SLAM algorithm can exhibit strong robustness when observing distinctive features. However, some environments, such as tunnels or empty rooms, may lack sufficient features to navigate reliably. This is called environmental degradation, which refers to the difficulties encountered in positioning and mapping in specific scenarios and then lead to poor system performance. We propose an improved algorithm based on Non-Rigid Structure-from-Motion (NRSfM) and Deformable SLAM (DefSLAM). The method proposed in this paper can be used to deal with the degraded environment of visual sparse and repetitive features. In the experiment, the algorithm processes the close-up sequence of the degraded scene, both in the laboratory-controlled experiment and in the real-world sequence, to generate a 3D model of the scene relative to the mobile camera. It has been proved that the algorithm proposed in this paper achieves impressive performance on the dataset and the estimation method is robust to degradation.