Research on Improving the Robustness and Positioning Accuracy of Visual SLAM Based on Point-Line Feature Matching

In order to solve the problem of the failure of SLAM algorithm due to insufficient feature information of a single point in low texture environments, an indoor visual/inertial localization algorithm combining point-line feature matching is proposed. By optimizing the LSD line feature extraction algorithm, the close fusion of vision and IMU data is achieved, and an adaptation factor is introduced in the optimization objective to reduce the processing time. This method reduces the absolute trajectory error and improves the positioning accuracy and system robustness.