Visibility of Point Cloud

最新推荐文章于 2025-12-03 15:22:15 发布

原创

最新推荐文章于 2025-12-03 15:22:15 发布 · 856 阅读

0 ·

CC 4.0 BY-SA版权

文章标签：

#计算机视觉 #PCL #点云库

本文探讨了点云可见性分析中的一种常见方法——隐藏点移除(HPR)及其在噪声或非均匀采样点云中使用时的局限性。提到了Katz的方法可能存在假阳性问题，并介绍了其他替代方法，如基于表面三角形和体素的技术。目前作者正在实现光线投射和扫描线方法，并详细描述了这两个方法的基本思想。还提到已修改python-pcl库以从网格生成更密集的点云，并获取点的法线信息。代码支持CPU运行，但其他部分需要CUDA加速。

Currently, the method of hidden point removal HPR (Katz et al., 2007) is widely applied for the visibility analysis. The advantage of this technique is to avoid creating a surface from the point cloud which might be expensive and this led to analyze visibility efficiently with both sparse and dense clouds. However, when the point cloud is noisy or non-uniformly sampled, a robust HPR operator (RHPR) is preferred to be used (Mehra et al., 2010) to deal with these cases.

However, the Katz’s method has false positives. There are several other methods to detect the hidden points:
* The surface triangle based methods (the normal direction testing, triangle – ray intersection, Z – buffering method)
* The voxel based techniques(voxel-ray intersection, ray tracing and Z-buffering method)
* The hidden point removal HPR

I am currently implementing the ray tracing (in an image spacing) and scan line. The idea is relatively the same except that the ray tracing(cas