【摘要】 针对特征点法在非结构化复杂场景下没有明显特征导致特征提取失败的问题，提出了一种基于布料算法特征提取的改进激光SLAM算法。首先，提出基于布料算法提取垂直方向上的表面点云为特征点的新型特征提取方法，其能有效反应点云周围结构的垂直形状，且包含更多的全局点云信息；其次，采用3DSC方法作为特征点的描述符，确定相邻点云的对应点，利用改进的RANSAC方法进行粗配准，结合ICP方法进行精配准，完成位姿估计，提高配准精度。最后，采用全局描述符LiDAR-Iris进行回环检测，减少累计误差。实验结果表明，该方法在精度、低漂移、稳定性方面优于A-LOAM、LeGO-LOAM和LIOSAM，提升了系统的整体性能。更多还原

【Abstract】 Aiming at the problem that the feature point method has no obvious features in unstructured and complex scenes, which leads to the failure of feature extraction, an improved laser SLAM algorithm based on the feature extraction of cloth algorithm is proposed. Firstly, a new feature extraction method based on the cloth algorithm is proposed to extract the surface point cloud in the vertical direction as feature points, which can effectively reflect the vertical shape of the structure around the point cloud and contain more global point cloud information.Secondly, the 3DSC method is adopted as the descriptor of the feature points to determine the corresponding points of the neighbouring point clouds, and the improved RANSAC method is used to perform the coarse alignment, which is combined with the ICP method to perform the fine alignment to complete the pose estimation. The improved RANSAC method is used for coarse alignment and combined with the ICP method for fine alignment to complete the pose estimation and improve the alignment accuracy. Finally, the global descriptor LiDAR-Iris is used for loopback detection to reduce the cumulative error. The experimental results show that the proposed method is better than A-LOAM and LeGO-LOAM in terms of accuracy, low drift and stability, and improves the overall performance of the system.更多还原