DymSLAM: 4D Dynamic Scene Reconstruction Based on Geometrical Motion Segmentation

Most SLAM (Simultaneous Localization and Mapping) algorithms are based on the assumption that scene is static. However, in practice, most real scenes usually contain moving objects. In this letter, we introduce DymSLAM, a dynamic stereo visual system being capable of reconstructing 4D (3D + time) with rigid Unlike previous attempts have considered objects as outliers ignored them, DymSLAM obtains 6DoF motion trajectory 3D models about We segment different by multi-motion segmentation approach obtain accurate masks Besides ego-motion, our can model object global reference frame while simultaneously dense map static background. Meanwhile, does not rely semantic cues or prior knowledge suitable for unknown conducted experiments real-world indoor environment where both camera were wide range. The results proved proposed method state-of-the-art use environment.