【摘要】 针对弱纹理环境下SLAM系统只依靠单一传感器鲁棒性较差的问题，提出一种视觉、单线激光雷达与惯性相结合的机器人SLAM算法。在视觉与雷达预处理阶段，视觉提取点线特征，同时雷达帧间匹配过程采用激光点到其最近两个点连线的距离构建误差方程，实现更高精度匹配效果。采用惯性传感器与轮速里程计进行雷达运动畸变校正，同时雷达估计信息为单目点线特征三角化提供良好深度值，再利用点线视觉信息、雷达点云信息与惯性测量单元紧耦合优化机制提高机器人SLAM的精确度。最后，将该方法在仿真环境和真实弱纹理环境进行实验。结果表明：该方法定位准确率达到98.6%,在弱纹理环境中定位效果具有较强的鲁棒性和准确性，满足实际需求。更多还原

【Abstract】 Aiming at the poor robustness that SLAM system only depends on a single sensor in weak texture environment, a robot SLAM algorithm combining vision, single-line lidar and inertia was proposed.In the pre-processing stage of vision and radar, the feature of point and line was extracted by vision, and the error equation was constructed by the distance between the laser point and its nearest two points in the matching process of radar frames, so as to achieve higher precision matching effect.Inertial sensor and wheel speed odometer were used to correct radar motion distortion, and radar estimation information provided good depth value for triangulation of monocular point-line features.Then, the precision of robot SLAM was improved by using the close coupling optimization mechanism of point-line visual information, radar point cloud information and inertial measurement unit.Finally, the method was tested in simulation environment and real weak texture environment.The results show that the positioning accuracy of this method reaches 98.6%,and the positioning effect in weak texture environment is robust and accurate, which meets the actual needs.更多还原