This software is part of the perception lessons in the Udacity Robotics Nanodegree Program. Sensorstick performs object segmentation on 3D point cloud data using python-pcl
based on the Point Cloud Library.
The steps are:
-
Load a point cloud file example -- the tabletop, with objects on it.
-
Perform Voxel Grid Downsampling to reduce the image size.
-
Implement a Pass Through Filter.
-
Perform RANSAC plane segmentation.
-
Extract inliers - the table.
-
Extract outliers - the objects on the table.
-
Use the Euclidean distance to find clusters (individual objects).
-
Color the individual objects.
-
Publish the tabletop and objects as ROS messages to the Gazebo simulation.
These steps are implemented as a ROS node using the publisher/subscriber method.