基于三维激光扫描的岩质边坡结构面自动识别方法

The improved automatic identification method of rock slope structural plane based on 3D laser scanning

针对解决传统结构面调查效率低,而现有结构面自动识别技术多为非监督式聚类或夹角阈值分割,操作者主观影响过大或者鲁棒性较差都可能导致优势结构面数量判定不准确的问题,本文提出了一种基于改进分水岭的结构面识别方法。首先,以结构面点云为数据源,采用PCA主成分分析计算平面法向量;随后,求解出点云局域平面的产状并使用高斯核密度估计分析点云分布密度;最后,利用改进分水岭算法实现岩体优势结构面自动聚类。结果表明:本文采用的密度簇阈值分割算法,无需确定种数类别,可直接得到结构面最优聚类组数及产状。算法识别出的结构面产状误差最大为2.9°,满足精度要求,为岩体结构面精细化识别提供了可靠保证。

Most of the existing automatic watershed identification techniques,such as unsupervised clustering or angle threshold segmentation,have caused excessive subjective influence of operators or poor robustness,which leads to the determination of the number of dominant structural planes are not accurate.In this paper,an improved watershed based structural plane identification method is proposed to solve the problem of traditional structural plane survey efficiency.Firstly,structural point clouds were used as data sources and PCA was used to calculate the plane normal vector.Then,the occurrence of the local plane of the point cloud was solved and the distribution density of the point cloud was analyzed by using the Gaussian Kernel density estimation.Finally,the improved watershed algorithm was used to realize the automatic clustering of the dominant structural plane of rock mass.The results show that the density cluster threshold segmentation algorithm adopted in this paper can directly obtain the optimal cluster number and occurrence of structural plane without determining the number category.The maximum occurrence error of structural plane identified by the algorithm is 2.9°,which meets the accuracy requirements.Therefore,the improved method provides a reliable guarantee for fine identification of structural plane of rock mass.