How to rasterize non-floor point clouds

I am planning to use the artificial potential field method to “partially avoid obstacles”. My scene is relatively simple, consisting of a flat ground and some walls, and there will be some irregular obstacles in the scene (such as school bags placed on the ground).

I plan to first remove the floor from the point cloud data, and then perform two-dimensional gridding. Finally, I will calculate the potential field of these gridded data seat obstacles.

But now I have some problems:

How to remove ground from a point cloud? I am currently using camera.retrieveMeasure(points, MEASURE::XYZ, memory_type, image_size) to obtain real-time point cloud data. I want to know how to remove the floor from the point cloud data?

My Mat is currently processed directly on the GPU. How can I convert these processed point clouds into two-dimensional grids in the fastest and least CPU intensive way?

Hi @yuanlu
for this type of processing, I suggest you look at external point cloud libraries, i.e. PCL and Open3D.

1 Like