Monocular Visual Inertial SLAM Algorithm Based on Dynamic Fusion of Point-line Features
For the problems of poor robustness and low localization accuracy of traditional visual SLAM(Simultaneous Localization and Mapping)algorithm in dark texture environment,a monocular visual inertial SLAM algorithm based on dynamic point-line coupling is given in this paper.The dynamic threshold value of extracted line features is set to improve the utilization of line features,then the simi-larity of point and line feature positions is used to extract new point-line coupled features and construct a point-line coupled residual model,and the point-line coupled residuals are used to search the relationship between points and line features,and finally the point-line coupled residuals are integrated into the back-end sliding window optimization to construct a minimization cost function.The con-structed algorithm is simulated on the EuRoC public dataset,and the experimental results show that the designed algorithm reduces the localization error significantly compared with the VINS-Mono algorithm and the PL-VIO algorithm,which effectively enhances the ro-bustness and improves the localization accuracy of the system.
NETL
NSTL
万方数据
