RGB-D Based Visual SLAM Algorithm for Indoor Crowd Environment

• DOI: 10.1007/s10846-023-02046-3
• Publication Date: 2024-02-02T09:02:30Z
• Authors: Jianfeng Li; Juan Dai; Zhong Su; Cui Zhu
• Keywords: RGB color model

Abstract Most current research on dynamic visual Simultaneous Localization and Mapping (SLAM) systems focuses scenes where static objects occupy most of the environment. However, in densely populated indoor environments, movement crowd can lead to loss feature information, thereby diminishing system’s robustness accuracy. This paper proposes a SLAM algorithm for dense environments based combination ORB-SLAM2 framework RGB-D cameras. Firstly, we introduced dedicated target detection network thread improved performance network, enhancing its coverage crowded resulting 41.5% increase average Additionally, found that some points other than humans box were mistakenly deleted. Therefore, proposed an standard deviation fitting effectively filter out features. Finally, our system is evaluated TUM Bonn datasets compared with state-of-the-art methods. The results indicate pose estimation error reduced by at least 93.60% 97.11% high dataset, respectively. Our method demonstrates comparable recent