Infrared visual-inertia-UWB fusion positioning method for drones in cabin environments

In cabin environments, challenges such as the lack of global navigation satellite system (GNSS) signals and low light conditions pose difficulties for the positioning of inspection drones. Additionally, positioning methods based on autonomous sensors, such as vision, LiDAR, and inertial systems, will produce large cumulative errors during long-term operation in the absence of loop closure detection and global pose graph optimization. To address these issues, a double layer sliding window factor graph fusion positioning method based on infrared (IR) vision images, inertial measurements, and ultra-wideband (UWB) ranging information is proposed. Firstly, an IR vision camera and a Contrast Limited Adaptive Histogram Equalization (CLAHE) method are introduced to enhance the drone's perception of the cabin environment, while the SuperPoint network is utilized to extract stable feature points. Then, a factor graph optimization approach is employed to fuse the information of visual, inertial, and UWB sensors, leveraging the drift-free characteristics of UWB measurements to reduce cumulative errors in the system. Finally, the effectiveness of the proposed method was verified through simulation experiment. Experimental result demonstrates that compared with the traditional visual inertial fusion positioning method, the positioning accuracy of the proposed approach is improved by 59.56%.