An Indoor Positioning Method Leveraging Geomagnetism/IMU/Vision

Image retrieval-based visual positioning demonstrates significant potential in indoor positioning. However, it is susceptible to locally repetitive structures and sparsetexture surfaces. To address this problem, we propose a method based on multiple sensors embedded in a smartphone, which consists of two stages. In the visual feature optimization stage, we introduce a multiple visual cue-guided feature optimization (MCF) method. MCF jointly leverages multiple visual cues to guide the feature optimization process, thereby enhancing the discriminability of visual features. In the fusion stage, we propose a fusion method incorporating inertial information assistance and an evidence-driven (IA-ED) process, reducing visual errors that still exist at very few positioning points caused by locally repetitive structures and sparse-texture surfaces. First, we use inertial information to assist single-sensor positioning based on vision and geomagnetism. Then, multiple trajectories and sequential features are extracted from both vision and geomagnetism, and fed into an evidence-driven process for comprehensive analysis, thereby achieving the fusion of vision and geomagnetism. Subsequently, particle filtering is employed to further fuse the fusion result with pedestrian dead reckoning (PDR) to estimate the final position. The performance of the method was evaluated in real indoor environments. The experimental results show that the proposed method can achieve accurate and continuous positioning.