Pointcloud PCD of Stronghold Tower?

[joshumax]

Hi,

I'm a student currently working on the S.L.A.M. guidance/shooting code for my team's robot and recently got a port of the Pointcloud library working on our NVIDIA TK1 vision host computer. For goal detection, we're using an Orbbec Astra Pro Custom Edition structured light depth sensor in conjunction with OpenNI, CUDA-CV and PCLs Euclidean Cluster Extraction. My question is, has anyone built a replica of the tower? If so, is there anywhere I might be able to find a PCD dump of the tower face-on taken from another structured light sensor? Having the PCD on hand would make development and testing far easier than it currently is for me.

Thanks!

[marshall]

This is really cool stuff! I wish I could help with getting the data you need but I'm just thrilled to hear of another team working on SLAM. I hope you get some recognition for it at a competition because this is awesome!

[TheOtherGuy]

I would be shocked if someone had a point cloud already. Perhaps getting vertices from a mesh file is enough? I'm not sure if you're acquainted with CAD tools, but here's a STEP model of the field on GrabCAD you could work with:
https://grabcad.com/library/frc-field-2016-2

If that's not good enough, and you have a lot of time on your hands, you could install FRC Sim and figure out how to add a Kinect to the default robot.

https://wpilib.screenstepslive.com/s/4485/m/23353
http://gazebosim.org/tutorials?tut=ros_gzplugins#OpenniKinect

[joshumax]

I would be shocked if someone had a point cloud already. Perhaps getting vertices from a mesh file is enough? I'm not sure if you're acquainted with CAD tools, but here's a STEP model of the field on GrabCAD you could work with:
https://grabcad.com/library/frc-field-2016-2

If that's not good enough, and you have a lot of time on your hands, you could install FRC Sim and figure out how to add a Kinect to the default robot.

https://wpilib.screenstepslive.com/s/4485/m/23353
http://gazebosim.org/tutorials?tut=ros_gzplugins#OpenniKinect

I should've noted that for the time being I _was_ using a point cloud database created from a model of the tower made in Blender. The problem is, the point cloud is _too_ good, when in reality things like the retro-reflective tape near the goal causes incorrect readings from the structured light sensor (or at least on the Orbbec Astra, which uses a strange diffraction grating). I was also using Gazebosim with ROS for a short period of time, but designing the robot and configuring it to work in Gazebo started to take too long :(