search

xingxuejun1989 / SCVC

8 stars 1 forks source link

Does the code include feature matching and validation? #1

Open Zjsik opened 4 months ago

Zjsik commented 4 months ago

Hello, I would like to use your code as part of my acting thesis and would like to ask you a few questions. May I ask if the code is suitable for registration of LiDAR point clouds? What do you usually need to obtain to capture the viewpoint of a point cloud in the main code? Is there a new feature matching method mentioned in the paper included in the code? I hope you can reply to me, thank you.

xingxuejun1989 commented 4 months ago

Thanks, I really appreciate your attention to our proposed method. (1) SCVC is very suitable for the registration of LiDAR point clouds. (2) For the initial captured point cloud without transformation, the viewpoint is (0, 0, 0), which is the position coordinate of the LiDAR device. If it has been transformed, you need to manually mark the viewpoint. (3) The code contains feature matching code, but the matching method is not optimal and there is still a lot of room for improvement.

Zjsik commented 4 months ago

May I ask if the code for feature matching is the kd tree part? And when I run CMake, I will encounter the error "CMake Error: CMake can't terminate linker language for target: kd_tree". How can I resolve this issue?

xingxuejun1989 commented 4 months ago

Thank you for finding the problem. I have resubmitted the code.

Zjsik commented 3 months ago

Hello, may I ask what information is required in the code to obtain point cloud data for the sentence "std:: string postpath="../data/KITTI/icp "? Is it ground truth data (i.e. paired transformation matrix)?

xingxuejun1989 commented 3 months ago

You should be able to download the experimental data and set the relevant parameters according to the address provided. It should be noted that the pose transformation matrix of KITTI is not accurate. I processed and verified the original pose data through ICP iterative optimization, which can also be downloaded and applied.

Zjsik commented 3 months ago

May I ask what information should be inputted into PostPath if I want to run the algorithm on my own dataset? I have used some publicly available TLS datasets that provide point clouds and Digital Terrain Models (DTM). Can I use them? If that's not possible, how should I get the information that PostPath should input?

xingxuejun1989 commented 3 months ago

It would be best to write the corresponding interface based on the algorithm code to test your data or provide the corresponding data according to the data format in the code.

Zjsik commented 3 months ago

Can you explain it more clearly? Is the input in std:: string postpath=the pose information of the point cloud?

Zjsik commented 3 months ago

Can you explain it more clearly? Is the input in std:: string postpath=the pose information of the point cloud?

xingxuejun1989 commented 3 months ago

std:: string postpath= The file path to save the 4*4 transformation matrix. The file name is "point cloud class number-the first point cloud number-the second point cloud number.pny".

Zjsik commented 3 months ago

Okay, thank you very much for your help.

Zjsik commented 3 months ago

Is the conversion validation function VDIR mentioned in the article included in the code?