Split data from lidar

Adreas_Papandreou · October 11, 2019, 9:04am

Hello,

Is it possible to split the lidar data by scan lines? For example i would like to process separately data that exist in the same scan (laser) line and create some metadata.

Thanks in advance.

holger-motional · October 11, 2019, 9:20am

Yes, we call the scan line ring index. It’s in the pointcloud but dropped by default:

github.com

nutonomy/nuscenes-devkit/blob/master/python-sdk/nuscenes/utils/data_classes.py#L247


def nbr_dims() -> int:
    """
    Returns the number of dimensions.
    :return: Number of dimensions.
    """
    return 4


@classmethod
def from_file(cls, file_name: str) -> 'LidarPointCloud':
    """
    Loads LIDAR data from binary numpy format. Data is stored as (x, y, z, intensity, ring index).
    :param file_name: Path of the pointcloud file on disk.
    :return: LidarPointCloud instance (x, y, z, intensity).
    """


    assert file_name.endswith('.bin'), 'Unsupported filetype {}'.format(file_name)


    scan = np.fromfile(file_name, dtype=np.float32)
    points = scan.reshape((-1, 5))[:, :cls.nbr_dims()]
    return cls(points.T)

Adreas_Papandreou · October 16, 2019, 11:01am

Thanks for your answer. I run the function from_file in data_classes.py and I saw that the value of ring_index does not exist everywhere. But wherever it exists, it has always the same value, which is 1. The image I uploaded shows the results for one random lidar file and the columns follow the order (x, y, z, intensity, ring index).

lidar_example

holger-motional · October 17, 2019, 4:05pm

How did you get these results? I picked the first lidar pointcloud and saw meaningful results:

import os
import numpy as np
from nuscenes.utils.data_classes import LidarPointCloud
sd = nusc.get(‘sample_data’, nusc.sample[0][‘data’][‘LIDAR_TOP’])
pcl_path = os.path.join(nusc.dataroot, sd[‘filename’])
scan = np.fromfile(pcl_path, dtype=np.float32)
points = scan.reshape((-1, 5))

points[:, -2]

array([ 4., 1., 2., …, 80., 75., 40.], dtype=float32)

points[:, -1]

array([ 0., 1., 2., …, 29., 30., 31.], dtype=float32)

So ring index really just increments by 1 each time, whereas intensity are meaningful int8 values.