PointCloud (Lidar) to RangeImage - Algorithm not using spherical model?

classic Classic list List threaded Threaded
2 messages Options
tbb
Reply | Threaded
Open this post in threaded view
|

PointCloud (Lidar) to RangeImage - Algorithm not using spherical model?

tbb
Hey,

quick question, I've been
digging around the pcl spherical range image to understand
how the PCL algo works and I've noticed the following equations in the getImagePoint function:

angle_x = atan2(point[0], point[1])
angle_y = asin(point[1], range)


to calculate the angles. Seeing these equations, I would assume
this algo is using a cylindrical model and not a spherical one (as it has been described
in the pcl documentation).

The theoretical model would be: (from papers) 
Cylindrical model: theta = atan2(x,y), alpha = arcsin(z / range)
Spherical model: theta = arcsin(y/sqrt(x^2+y^2), alpha=arcsin(z/range)

Maybe anybody can chime in and knows a bit more about this... :)

Thanks!


_______________________________________________
[hidden email] / http://pointclouds.org
http://pointclouds.org/mailman/listinfo/pcl-users
Reply | Threaded
Open this post in threaded view
|

Re: PointCloud (Lidar) to RangeImage - Algorithm not using spherical model?

Sérgio Agostinho

Unfortunately I don't think the original author Bastian Steder is not around to step in and the original commit was prior to the GitHub migration so I can't trace the original pull request to get more info.

The theoretical model would be: (from papers) 
Cylindrical model: theta = atan2(x,y), alpha = arcsin(z / range)
Spherical model: theta = arcsin(y/sqrt(x^2+y^2), alpha=arcsin(z/range)

Which papers exactly? The thing is that there are a number of spherical and cylindrical  projections. Which one do you think the actual code is or isn't implementing?

Cheers



_______________________________________________
[hidden email] / http://pointclouds.org
http://pointclouds.org/mailman/listinfo/pcl-users

signature.asc (836 bytes) Download Attachment