Distorted LiDAR Point Cloud

Hi,

I’m working on a pedestrian recognition project and am unsure about how to ‘undistort’ the LiDAR point clouds (see below, note the ‘parabolic’ points which make up the building on the right).

Will this present issues going forward? My main concern is properly aligning the annotation bounding boxes; will the bboxes share this distortion such that composite points can still be accurately identified?

Even if it won’t cause issues, not having the points aligned in visualizations is bothering me :sweat_smile: I’d assume any fixes would also have to be applied to the bbox coordinates in this case.

Thanks

Relevant Code

# Import nuScenes data
nusc = NuScenes(version='v1.0-mini', dataroot='data/mini/', verbose=False)

# Gather scene, sample, and sample data
scene = nusc.scene[2]
sample = nusc.get('sample', scene['first_sample_token'])  # First sample from the scene
sample_data = nusc.get('sample_data', sample['data']['LIDAR_TOP'])

# Get LiDAR data points
point_cloud_path = 'data/mini/' + sample_data['filename']
point_cloud = LidarPointCloud.from_file(point_cloud_path)

# Rotate and transform point cloud
cs_record = nusc.get('calibrated_sensor', sample_data['calibrated_sensor_token'])
point_cloud.rotate(Quaternion(cs_record['rotation']).rotation_matrix)
point_cloud.translate(np.array(cs_record['translation']))

Update

I have been tinkering with points_in_box() to identify points which make up individual objects, and the raw point cloud data is working without any issues. (the last block of code in the original post prevents the points from being discovered)

I would still be interested in working out how to remove the distortion for visualization purposes if anyone has a solution.

Hi. Can you elaborate what distortion you are referring to?

  • We perform motion compensation on the pointcloud, so regardless of the vehicle speed, there should not be any deformations of the pointcloud as a whole.
  • We already undistorted and rectified the images. So you can use them as is for fusion or to project points into the image.
1 Like

I think distorted was the wrong word to use. In the image in the original post, see how the points curve upwards away from the centre point.

This is my first time working with LiDAR data, is this normal? It doesn’t seem to be causing any issues with the stuff I’m writing. Is it something to do with the reflectivity of the points as they get further from the sensor?

Hi. I think it just looks confusing. If there is an obstacle close to the car the lidar rings will form these patterns. I don’t think there is any distortion.

1 Like

Alright I think I understand it now, thank you for clearing that up :smile: