I have a lidar data collected with Velodyne-128 in .bin format. I need to convert it into pcd format. I use NVIDIA Driveworks for data manipulation but there is no tool to convert lidar binary data into pcd.
Thus, is there a method to convert binary lidar data into pcd format?
Here is a snippet to converting a lidar data (in .bin format) into .pcd format
with open ("lidar_velodyne64.bin", "rb") as f:
byte = f.read(size_float*4)
while byte:
x,y,z,intensity = struct.unpack("ffff", byte)
list_pcd.append([x, y, z])
byte = f.read(size_float*4)
np_pcd = np.asarray(list_pcd)
pcd = o3d.geometry.PointCloud()
v3d = o3d.utility.Vector3dVector
pcd.points = v3d(np_pcd)
According to the example here, you can save it to file it as below:
import open3d as o3d
o3d.io.write_point_cloud("copy_of_fragment.pcd", pcd)
Although all the other solutions are probably correct, I was looking for a simple numpy-only version. Here it is:
import numpy as np
import open3d as o3d
# Load binary point cloud
bin_pcd = np.fromfile("lidar_velodyne64.bin", dtype=np.float32)
# Reshape and drop reflection values
points = bin_pcd.reshape((-1, 4))[:, 0:3]
# Convert to Open3D point cloud
o3d_pcd = o3d.geometry.PointCloud(o3d.utility.Vector3dVector(points))
# Save to whatever format you like
o3d.io.write_point_cloud("pointcloud.pcd", o3d_pcd)
I found a code from github (reference is given below):
import numpy as np
import struct
from open3d import *
def convert_kitti_bin_to_pcd(binFilePath):
size_float = 4
list_pcd = []
with open(binFilePath, "rb") as f:
byte = f.read(size_float * 4)
while byte:
x, y, z, intensity = struct.unpack("ffff", byte)
list_pcd.append([x, y, z])
byte = f.read(size_float * 4)
np_pcd = np.asarray(list_pcd)
pcd = geometry.PointCloud()
pcd.points = utility.Vector3dVector(np_pcd)
return pcd
