LAUSR

Simultaneous localization and mapping (SLAM) is considered to be indispensable for intelligent robots. However, most existing SLAM systems operate under the assumption of static environment. These systems often falter in the presence of moving objects, significantly reducing their efficacy in real-world applications. In this article, an RGB-D SLAM system for indoor dynamic environments is introduced, which aims to accurately estimate the camera pose and reconstruct static environments. First, a statistics-based dynamic detection method is proposed; it models the feature points by two geometric constraints and achieves adaptive detection through a chi-square test. Then, a pose optimization method based on static weight is established to enhance pose estimation accuracy. Besides, dynamic regions are effectively identified through a multimask fusion algorithm, which fuses multiple masks from geometric, semantic, and depth information to ensure the reliable construction of static scenes. The proposed method is evaluated using public datasets and real-world scenarios, demonstrating its competitive performance relative to state-of-the-art methods and its adaptability to different dynamic environments.