https://github.com/willguimont/r2_lidar_icp

https://github.com/willguimont/r2_lidar_icp

icp lidar lidar-point-cloud machine-learning point-cloud point-cloud-registration

Last synced: 3 months ago
JSON representation

• Host: GitHub
• URL: https://github.com/willguimont/r2_lidar_icp
• Owner: willGuimont
• License: mit
• Created: 2021-10-28T22:04:47.000Z (over 3 years ago)
• Default Branch: master
• Last Pushed: 2024-11-18T19:42:22.000Z (5 months ago)
• Last Synced: 2024-11-18T20:44:04.585Z (5 months ago)
• Topics: icp, lidar, lidar-point-cloud, machine-learning, point-cloud, point-cloud-registration
• Language: Python
• Homepage:
• Size: 16.7 MB
• Stars: 3
• Watchers: 2
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README.md
  • License: LICENSE

Awesome Lists containing this project

README

        
# r2_lidar_icp

Iterative Closest Point (ICP) algorithm for lidar odometry.

r2_lidar_icp is a simple Python implementation of the ICP algorithm aimed to be easy to understand and modify.

[r2_lidar_icp in action](https://youtu.be/9I7yZk28Vi0?si=otcAcv2YrVtqMob7)

## Pipeline for ICP

- Preprocess (downsample, filter, compute descriptors, etc.)

- While not converged

  - Find matches

  - Reject outlier matches

  - Compute transformation

  - Transform point cloud

- Return transformation

## TODO

- [ ] Add tests

- [ ] Add 3D support for PointToPlaneMinimizer

- [ ] Profile and optimize (numba, cython, etc.)

- [ ] Refactor PointCloud to contain its descriptors

- [ ] Robust kernel

- [ ] De-skewing using IMU

- [ ] Dynamic object detection

- [ ] Diffentiable ICP

- [ ] Differentiable LiDAR simulator

- [ ] More filters

  - [libpointmatcher filters](https://libpointmatcher.readthedocs.io/en/latest/DataFilters/#filter-index)

  - [libpointmatcher config](https://libpointmatcher.readthedocs.io/en/latest/Configuration/)