由于不同点云数据的曲率变化不同，导致难以计算关键特征的尺度参量，配准后源点云和目标点云在相同位置处，两者在三维空间中的距离存在差异，即配准后点云之间不重合、存在落差，配准精度较低。为此，提出基于BIM模型的三维激光点云粗配准算法研究。通过构造邻域超平面，计算关键点的特征尺度，并采用正则化系数将曲率归一化至相同水平，根据一致的曲率水平集中求取关键特征的尺度参量，由此提取关键特征点，结合点云数据和BIM模型的四平面基，计算两者的补丁数量，引入似真度函数估计点云的粗配准初始位置，以此为依据，计算初始变换矩阵的平移向量，结合收敛条件得到配准过程的最佳旋转矩阵，由此完成源点云数据和目标点云的粗配准。实验结果表明，利用所提方法进行三维激光点云粗配准，目标点云与源点云完全重合，整体配准效果较好；相同位置处的云落差更小，最小点云对应云落差值为0.004。

Because of the different curvature changes of different point cloud data, it is difficult to calculate the scale parameters of key features.After registration, the source point cloud and target point cloud at the same position, the distance between them in 3D space is different, that is,after registration, the point clouds do not overlap, there is a gap, and the registration accuracy is low. Therefore, we proposed a rough registration algorithm for 3D laser point clouds based on BIM model. By constructing a neighborhood hyperplane, we calculated the feature scale of key points, and used regularization coefficient to normalize the curvature to the same level. According to the consistent curvature level, we obtained the scale parameters of key features, and extracted the key feature points. Combining the point cloud data with the 4-plane basis of BIM model,we calculated the patch numbers of feature points, and introduced the approximate truth function to estimate the initial position of rough registration of point clouds. On the basis of this, we calculated the translation vector of the initial transformation matrix, and obtained the optimal rotation matrix in the registration process by combining the convergence conditions. The experimental results show that using the proposed method for rough registration of 3D laser point cloud, the target point cloud and the source point cloud completely coincide, and the overall registration effect is good. The cloud drop at the same position is smaller, and the cloud drop of minimum point cloud is 0.004.