Post-processing of Kitti dataset, bin, and pcd data

Question

The .bin file or pcd file of the Kitti dataset has 64 layers.

But I will do 3D detection with my 3 VLP16 Lidars.

• For learning, I want to reduce the number of 64 layers in the Kitti dataset to 16 or 32.

The kitti dataset consists of a .bin file, which can be converted to a .pcd file.

Can I post-process these files in my desired direction?

Answer 1

Yes, you can do it. The approach is using range image. First, you need to read the pcd file line-by-line. Then you calculate the angle of those pointcloud and save it in a form of 2D image with numpy. Create new numpy array, then select a row you want to use (e.g. from 64 channel->16 channel you will use row number: 0, 4, 8, etc). Then convert back to pointcloud data.