search

tudelft3d / City3D

Large-scale LoD2 Building Reconstruction from Airborne LiDAR Point Clouds
https://github.com/tudelft3d/City3D
GNU General Public License v3.0
223 stars 36 forks source link

Point cloud data testing #44

Closed najsdh closed 1 month ago

najsdh commented 2 months ago

图片1 I would like to ask why such a point cloud input into the program crashed. I thought it was because the amount of data was too large, so I diluted the point cloud, but the program still crashed after entering the point cloud

LiangliangNan commented 2 months ago

By "entering the point cloud", did you mean the program crashes when loading the data? If so, can you post your point cloud here so we can debug it?

najsdh commented 2 months ago

plydata.zip Hello, this is the code I input for the test, please help me to see if there is any problem. If there is a problem, how can I adjust it so that it can reconstruct the model. These point cloud data are a building that I intercepted from a whole block of point cloud data, maybe the roof is noisy and irregular.

LiangliangNan commented 2 months ago

Your ZIP file contains multiple PLY fiels, which one should be used? Beside, you mentioned "this is the code", did you mean you make changes to the City3D code? If so, what changes did you make? Anyway, to understand what your issue is, please precisely describe how others can reproduce your issue.

najsdh commented 2 months ago

I'm sorry, I may have misstated that. What I want to say is that this is the point cloud data that I used for testing, and I seem to have failed to reconstruct every single one of them. I would also like to know how the height of the reconstructed building is obtained if only the point cloud data of the roof is tested.

LiangliangNan commented 2 months ago

I still don't understand what your issue was. You mentioned the program crashed and failed for all the test examples. Anyway, I quickly tested with your data and found the it worked well. Below are the snapshots for building 1 and building 2 (I didn't test with others because the first two were already successful).

Building 1: build1

Building 2: build2

The results seem quite good given the very low quality of the input point clouds. Besides, our method assumes the ground plane is at Z=0, but it seems your point clouds may have some aribitrary translations.

najsdh commented 1 month ago

Thank you. Maybe there's something wrong with the way I handled it. Let me check it again. Thank you very much for your help