search

hku-mars / HBA

[RAL 2023] A globally consistent LiDAR map optimization module
GNU General Public License v2.0
421 stars 58 forks source link

Questions about HBA results #33

Open aksrb1030 opened 4 months ago

aksrb1030 commented 4 months ago

Hi. Thanks for your great work! I tried to my own dataset but result something wrong! Sensor : Hesai XT32

max_iter = 30;
donwsample_size = 0.05;
voxel_szie = 2.5;
eigen_ratio = 0.05;
reject_ratio = 0.05;

Before, HBA Screenshot from 2024-04-11 11-41-47

After, HBA Screenshot from 2024-04-11 14-55-43

Please know how to soved!

samsdolphin commented 4 months ago

Hi @aksrb1030, thanks for using our work. From your screenshot, I believe the initial trajectory is of good quality and the parameter setting is fine, I think it might be due to some configuration issues. Maybe you could share your data with me, so I could try it with my laptop?

aksrb1030 commented 4 months ago

Sure!! I sent you google drive link your email to liuxiyuan95@gmail.com

phucvinh-nguyen commented 4 months ago

Hi @samsdolphin, i met the same problem, i realize that the x coordinate has changed sign after optimization. Please help me to solve it!

phucvinh-nguyen commented 4 months ago

Can i send my data to your email - liuxiyuan95@gmail.com?

tempted111 commented 3 months ago

Hi. Thanks for your great work! I tried to my own dataset but i do not konw how to get tx,ty,yz,qw. Are their values obtained through the mean of the point cloud？

samsdolphin commented 3 months ago

Hi. Thanks for your great work! I tried to my own dataset but i do not konw how to get tx,ty,yz,qw. Are their values obtained through the mean of the point cloud？

The pose of each LiDAR scan can be obtained from a general LIO system, so for example, you can use fastlio to get it.

tempted111 commented 3 months ago

Hi. Thanks for your great work! I tried to my own dataset but i do not konw how to get tx,ty,yz,qw. Are their values obtained through the mean of the point cloud？

The pose of each LiDAR scan can be obtained from a general LIO system, so for example, you can use fastlio to get it.

Thanks, Thanks for your great work! Our data has a pose every 8ms, which generates over 8000 PCD files. Therefore, I want to merge 15 poses to form a pose for 200ms, The number of PCD file files will be reduced to about 1000. How should I merge these 15 poses? I tried to take the average of these 15 poses before, but the result was poor. Can you give me some suggestions? Thank you

tempted111 commented 2 months ago

Hi. Thanks for your great work! I tried to my own dataset but result something wrong! Before HBA: Screenshot from 2024-06-14 18-02-26 After HBA: Screenshot from 2024-06-14 18-15-02 Please know how to soved!

tempted111 commented 2 months ago

I do not modify the parameter