Visual-Inertial SLAM Technology Based on Monocular Infrared Camera

Traditional visual-inertial odometry based on visible light cameras experiences significant visual degradation in complex lighting environments, posing significant challenges for the practical application of SLAM systems. Thermal infrared cameras are less affected by changes in lighting conditions and can operate around the clock. However, due to issues such as weak textures, low contrast, and low signal-to-noise ratio in infrared images, existing SLAM methods are not directly compatible with thermal infrared cameras. Edge features exhibit minimal performance degradation in infrared images. To address these challenges, we propose an infrared visual-inertial SLAM method based on point and line features. This method employs techniques such as bilateral filtering, stripe noise filtering, and adaptive histogram equalization to enhance image contrast and filter noise, overcoming the issues associated with poor-quality infrared images. Subsequently, an improved feature extraction method is employed to extract point and line features and track them. Finally, real-time state estimation is achieved by nonlinear optimization that minimizes reprojection errors over a sliding window.