Only released in EOL distros:  

Package LinksDependencies (11)

Package Summary

Released
Documented

Simple library for segmentation of tabletop and shelf surfaces

Contents

  1. Overview
  2. Demo
    1. surface_perception_demo
      1. Parameters

Use GitHub to report bugs or submit feature requests. [View active issues]

Overview

demo

surface_perception is a simple tabletop/shelf perception pipeline.

surface_perception::Segmentation is the main API. It takes in a point cloud, where the positive "z" direction points up. It also assumes that the point cloud has been cropped down to a tabletop/shelf scene.

Example: 

   1 surface_perception::Segmentation seg;
   2 seg.set_input_cloud(pcl_cloud);
   3 
   4 // (Optional) Set the particular indices to be used in the input point cloud,
   5 // for the segmentation operation. If the indices is not set, the entire
   6 // input cloud will be used.
   7 seg.set_indices(point_indices);
   8 
   9 // (Optional, default 10) Set the maximum difference between the normal
  10 // vector of surfaces and the z-axis.
  11 seg.set_horizontal_tolerance_degrees(10);
  12 
  13 // (Optional, default 0.005) Set the maximum distance between the surface
  14 // inliers and the plane that represents the surface.
  15 seg.set_margin_above_surface(0.01);
  16 
  17 // (Optional, default 0.01) Set the required distance between each cluster
  18 // corresponding to each object, during the object extraction.
  19 seg.set_cluster_distance(0.01);
  20 
  21 // (Optional, default 10) Set the minimum size of object clusters.
  22 seg.set_min_cluster_size(10);
  23 
  24 // (Optional, default 10000) Set the maximum size of object clusters.
  25 seg.set_max_cluster_size(10000);
  26 
  27 // (Optional, default 5000) Set the minimum number of inliers contained by a
  28 // surface.
  29 seg.set_min_surface_size(5000);
  30 
  31 std::vector<SurfaceObjects> surface_objects;
  32 bool success = seg.Segment(&surface_objects);

Demo

To run the demo: 

rosrun surface_perception demo target_point_cloud_frame cloud_in:=input_point_cloud_topic

If target_point_cloud_frame is not provided, the point cloud will be processed in base_link.

surface_perception::SurfaceViz visualizes the result: 

   1 SurfaceViz viz(marker_pub);
   2 viz.set_surface_objects(surface_objects);
   3 viz.Show();
   4 
   5 // Removes the visualization.
   6 viz.Hide();

surface_perception_demo

This node runs the demo of surface_perception

Parameters

crop_min_x (float, default: 0.0)
  • The minimum x coordinate of the cropped input cloud.
crop_min_y (float, default: -0.5)
  • The minimum y coordinate of the cropped input cloud.
crop_min_z (float, default: 0.05)
  • The minimum z coordinate of the cropped input cloud.
crop_max_x (float, default: 1.3)
  • The maximum x coordinate of the cropped input cloud.
crop_max_y (float, default: 0.5)
  • The maximum y coordinate of the cropped input cloud.
crop_max_z (float, default: 2.0)
  • The maximum z coordinate of the cropped input cloud.
horizontal_tolerance_degrees (float, default: 10.0)
  • The maximum degree difference between normal vector of a surface and the z-axis.
margin_above_surface (float, default: 0.015)
  • The the maximum distance between the surface and its inliers.
cluster_distance (float, default: 0.01)
  • The maximum cluster distance for object extraction.
min_cluster_size (integer, default: 100)
  • The minimum number of points for an extracted object.
max_cluster_size (integer, default: 5000)
  • The maximum number of points for an extracted object.
min_surface_size (integer, default: 5000)
  • The minimum number of points a surface must have.
min_surface_exploration_iteration (integer, default: 1000)
  • The minimum number of iteration that the surface exploration algorithm must go through.

Wiki: surface_perception (last edited 2018-03-16 00:40:26 by YuTangPeng)

Except where otherwise noted, the ROS wiki is licensed under the
Creative Commons Attribution 3.0