地面激光三维扫描中球面标靶自动检测方法

Sphere Targets Automatic Detection in 3D Terrestrial Laser Scanning

提出了一种从地面激光点云数据中提取球面标靶目标的有效方法。该方法首先在单站数据的栅格结构中利用邻域距离突变提取遮挡边界点,同时对其进行空间聚类;然后,利用随机采样一致性方法,在各聚类结点中的二维栅格结构中探测近圆结构,同时,根据点到扫描中心距离和扫描角差估算圆半径和圆的一致集数,在通过估算半径约束的圆形区域所对应的三维点集中探测球体模型;最后,通过预设球体半径和估算球面点数约束的球体模型,作为最终球面标靶模型。实验结果表明,该方法能够在1 min之内完成千万级点云数据中的球面标靶探测工作。

An efficient method for extracting spherical targets from point cloud in terrestrial laser scanning is proposed in this paper. The method is as follows, firstly occlusion boundary points are extracted via the distance mutation between the neighbors in the single station’s raster structure, and spatially clustered. Then, the random sample consensus method is used to detect near-circular structures in the two-dimensional raster structure for each clustering node, meanwhile, the near-circular structure’s radius and consensus set number are estimated according to the distance from the detected center point to scanner center and the scan angle interruption, and the spherical models are detected in point cloud subset whose circular region can pass radius and point number threshold value. Finally, these spherical models are chosen as the final results. The experimental results show that the proposed method can effectively detect spherical targets in more than 10 million point clouds within one minute.