Camera-Ground Geometry Enhanced Monocular Thermal–Inertial Odometry With Point-Line Fusion

Recently, thermal-inertial odometry (TIO) using in-frared cameras has garnered the attention of some scholars due to its all-weather working ability. However, infrared cameras, with high noise and poor imaging quality, seriously affecting the performance of TIO. This issue is particularly pronounced in the application of monocular infrared camera on unmanned ground vehicles (UGV). Thus, we propose a multi-information enhanced monocular TIO for UGV using point and line features based on the specific thermal imaging characteristics of infrared cameras. Firstly, a real-time TIO with point line combination is achieved using more advanced line feature extraction and tracking scheme. Secondly, a parallel front-end for ground point feature process is developed to overcome the scale degradation issue of monocular TIO using camera-ground constraints. Besides, the redundancy in ground line feature modeling was resolved based on the geometric relationship between the camera and the ground. Finally, a common back-end is developed to perform real-time optimization within a sliding window framework in-cluding IMU pre-integration factors, point feature reprojection factors, line feature reprojection factors, camera-ground con-straint factors, and ground line constraint factors. The proposed system was validated on real-world datasets and compared with the current state-of-the-art algorithm. Experimental results show that the proposed system exhibits superior performance and can achieve accurate scale estimation in monocular TIO under insuf-ficient excitation of UGV. The accuracy of MIE-TIO has been improved by 67.88%~96.73%, 64.89%~98.15%, 73.91%~98.78%, and 47.41%~95.59%, respectively, compared with VINS-Fusion, PL-VINS, EPLF-VINS, and POL-VIO. The dataset is available at https://github.com/Fireflyatcode/MIE-TIO.