This repository aims to be a short guide to map LiDAR point clouds to a camera image frame in the Kitti dataset.
Robotics-vision in a crucial task to endowing autonomous driving vehicles with environmental awareness. Without robust perception pipelines, it would be impossible to obtain reliable decision-making systems. This notebook aims to be a short guide and teach how to map a LiDAR point cloud to an image frame synchronize at the same timestamp.
The notebook is divided in the following sections:
- Loading images, point clouds and transformation matrices
- Transforming Point cloud to camera image frame
- Filtering the point cloud and aligning
- pykitti
- numpy
- cv2
- matplotlib
- open3d
If you want to use this Notebook, please download the Kitti dataset and arrange its files as show below.
├── 2011_09_26 │ ├── 2011_09_26_drive_0005_sync │ │ ├── image_00 │ │ │ ├── data [154 entries exceeds filelimit, not opening dir] │ │ │ └── timestamps.txt │ │ ├── image_01 │ │ │ ├── data [154 entries exceeds filelimit, not opening dir] │ │ │ └── timestamps.txt │ │ ├── image_02 │ │ │ ├── data [154 entries exceeds filelimit, not opening dir] │ │ │ └── timestamps.txt │ │ ├── image_03 │ │ │ ├── data [154 entries exceeds filelimit, not opening dir] │ │ │ └── timestamps.txt │ │ ├── oxts │ │ │ ├── data [154 entries exceeds filelimit, not opening dir] │ │ │ ├── dataformat.txt │ │ │ └── timestamps.txt │ │ └── velodyne_points │ │ ├── data [154 entries exceeds filelimit, not opening dir] │ │ ├── timestamps_end.txt │ │ ├── timestamps_start.txt │ │ └── timestamps.txt │ ├── calib_cam_to_cam.txt │ ├── calib_imu_to_velo.txt │ └── calib_velo_to_cam.txt ├── assets │ ├── final_point_cloud.png │ ├── point_cloud.png │ └── point_mapped.png ├── image_laser_fusion.ipynb ├── Raw_PointCloud_Totaltxt └── README.md
Special thanks to JohanSamir for providing this awesome notebook.