MIT-SPARK / TEASER-plusplus

A fast and robust point cloud registration library
MIT License
1.81k stars 344 forks source link

[QUESTION] test on my own datasets, registration failed #130

Closed MZener closed 6 months ago

MZener commented 2 years ago

Thanks for your work. I try the code(WinTeaser version) on my own datasets which is collected from the realsense device(D415), the FPFH features are used in the test, here the results, registration failed, can you tell me how to adapt the code on my own datasets, what's the main parameters which need to be adjusted for the point cloud data of different quality. thansk a lot.

168007826-8a7cfa1c-561c-489a-92f8-ccebed88f41d 168009259-f512c404-3ce6-4aa4-89e5-91bb21108b27 168009268-ec2ff1e0-e1f7-4dd8-9c2e-d8092840349c
jingnanshi commented 2 years ago

What feature extraction algorithm did you use?

MZener commented 2 years ago

the FPFH features are used. 微信截图_20220514194405

jingnanshi commented 2 years ago

Can you try registering your source with itself using FPFH features?

MZener commented 2 years ago

Can you try registering your source with itself using FPFH features?

I have tried two different way to register. (1) From source to itself, registration success, 捕获3 (2) from source to itself which is applied with a random rotation and translation matrix, registration failed, 捕获 捕获1

jingnanshi commented 2 years ago

Do you mind posting the entire code snippet you used for doing the self-registration experiment? My guess is the correspondence-matching part is causing this issue.

LimHyungTae commented 6 months ago

If there are no further questions, we will close the issue :) Thx