PS-VINS: A Visual–Inertial SLAM System With Pedestrian Gait and Structural Constraints Using Smartphone Sensors
Dayu Yan,Chuang Shi,Tuan Li,Haitao Jiang,Ming Xia,Qikai Guo,You Li
DOI: https://doi.org/10.1109/jsen.2024.3351757
IF: 4.3
2024-03-02
IEEE Sensors Journal
Abstract:Due to the low-quality sensors of consumer-grade devices, such as smartphones, traditional point-based visual–inertial simultaneous localization and mapping (SLAM) techniques applied to smartphones commonly suffer from low localization accuracy and potential tracking divergence. Incorporating more measurements can enhance localization performance and robustness, but the additional computational consumption may not satisfy the real-time requirements in mobile location-based services, such as pedestrian positioning. For mobile SLAM applications focused on pedestrians in human-made environments, this article proposes PS-VINS, a real-time visual–inertial SLAM system using smartphone sensors, exploiting additional motion constraints from pedestrian gait and geometric constraints from structural line features. To enhance the pose estimation accuracy of back-end optimization, we incorporate motion measurement models of step velocity and length by utilizing extracted pedestrian gait information. In addition, we establish vanishing point (VP) and structural camera rotation measurement models based on structural line features. To ensure high efficiency, we present efficient methods for gait information extraction and VP estimation. PS-VINS can further integrate global navigation satellite system (GNSS) output from smartphones to achieve global positioning in open outdoor environments. The field test results, including both outdoor and indoor environments, demonstrate that PS-VINS addresses the issue of drift error due to the low-quality smartphone sensors, achieving improved localization accuracy and efficiency compared to the state-of-the-art (SOTA) SLAM techniques.
engineering, electrical & electronic,instruments & instrumentation,physics, applied