INPUT DATA

• use the localization result in the submap of Laser T&R
• and the submap, and a raw scan of VLP16

PROCESS

• use the kdtree to remove the background, keep the dynamic ojects(people, cars, etc) on.
• Cluster on the objects.

Results

• laserScan on the subMap:

• Use Knn to get the outliers:

• the cluster results:

• Process:

PROBLEM

• There is problem in the clustering algorithm.
• Maybe the true EMST is really needed.
• Many of cluster methods are waiting for me!

Thanks

• Thanks to Oxford Robotics Institue

  Wang, Dominic Zeng, Ingmar Posner, and Paul Newman. "What could move? finding cars, pedestrians and bicyclists in 3d laser data." Robotics and Automation (ICRA), 2012 IEEE International Conference on. IEEE, 2012.

• and ETHZ_ASL's libs:

  • libpointmatcher: https://github.com/ethz-asl/libpointmatcher
  • libnabo: https://github.com/ethz-asl/libnabo