The integrity of the excavation tunnel point cloud depends on the effectiveness of registering laser scans at each station. However, the complex surface structure of the excavation tunnels, characterized by a lack of significant features, poses registration challenges. To address these obstacles, we introduce a classification mechanism to the neighbor reweighted local centroid (NRLC) algorithm, resulting in the robust NRLC (RNRLC) for extracting concave–convex feature points. We then propose a registration method based on concave–convex encoding for excavation tunnels. First, concave–convex points of the point clouds are extracted using the RNRLC method. Next, the sets of concave–convex feature points are clustered using Density-Based Spatial Clustering of Applications with Noise(DBSCAN), and the centroid of each set is calculated as a concave–convex feature point, significantly reducing the number of processing points. Descriptors are constructed using the Euclidean distance and concave–convex information of each feature point and its neighboring points. Corresponding point pairs are identified using edit distance, and the transformation matrix is computed by constructing congruent triangles with these pairs. Experiments conducted with different noise ratios and overlap rates demonstrate the method’s high accuracy and robustness in aligning excavation tunnel point clouds, outperforming other methods in these conditions.

中文翻译：

---

基于凹凸提取和编码的开挖隧道点云配准


开挖隧道点云的完整性取决于在每个站点配准激光扫描的有效性。然而，开挖隧道复杂的表面结构，缺乏重要的特征，这带来了配准挑战。为了解决这些障碍，我们在邻居重加权局部质心 （NRLC） 算法中引入了一种分类机制，从而产生了用于提取凹凸特征点的稳健 NRLC （RNRLC）。然后，我们提出了一种基于凹凸编码的开挖隧道配准方法。首先，使用 RNRLC 方法提取点云的凹凸点。接下来，使用基于密度的噪声应用程序空间聚类 （DBSCAN） 对凹凸特征点集进行聚类，并将每组特征点的质心计算为凹凸特征点，从而显著减少处理点的数量。描述符是使用每个特征点及其相邻点的欧几里得距离和凹凸信息构建的。使用编辑距离标识相应的点对，并通过使用这些对构建全等三角形来计算变换矩阵。在不同噪声比和重叠率下进行的实验表明，该方法在对准开挖隧道点云方面具有很高的准确性和鲁棒性，在这些条件下优于其他方法。