BaxterInterface/README.md

4.1 KiB

BaxterInterface

Bricks for an Rviz based interface for objective sending. Maintainer: Antoine Harlé antoine.harle@etu.upmc.fr

What is this ?

This is a ROS package implementing two working ROS modules : rviz_interface & asift_matching It also include other ongoing non-ROS modules, mostly for test purposes.

rviz_interface

This ROS modules implements markers and a panel for Rviz for sending StateSpace messages.

asift_matching

This ROS modules implements a wrapper for a SIFT based algorithm (Affine SIFT). It publish the position of a desired object, defined by its references, on a topic. Thus refrences are required for the module to work properly. It can be loaded through a text file or a series of images (which paths are defined in the launch file).

  • texte reference generation (BaxterInterface/asift_gen folder):
  • Add the images used for reference in the "references" folder.
  • open asift_match.cpp
  • Search at the beggining of the file the "REFERENCE DEFINITION"
  • In refData, enter the path/name of your reference images.
  • In nb_ref, enter the number of images.
  • cmake . && make
  • You can now run the program which is going to build the references and try to match with an image given in input : asift_match input.png
  • The generated text file will be located in the results folder.

Status

Done

Todo

asift_matching - ROS_matcher : Working loader of images for reference. asift_matching - ROS_matcher : Improve reactivity. Slow process due to the message queue length ? asift_matching - ROS_matcher : Swap the callback function to a service called for matching.

How to use it ?

Requirements

  • ROS Indigo (probably works in other versions too)
  • PCL
  • QT 5
  • Eigen 3

Usage

  • Clone the repository to your catkin src
  • catkin_make

Simple Rviz interface

  • Run : roslaunch rviz_interface interface.launch
  • Setup rviz :
  • set fixed frame to camera_rgb_optical_frame.
  • add an InteractiveMarker and set topic to (server_topic parameter)/update.
  • add a Marker and set topic to vizualization_topic parameter.
  • add the Rviz interface panel.

Rviz interface with simple moving tracker

  • Run :
  • roslaunch openni_launch openni.launch
  • roslaunch rviz_interface interface_plannar_seg.launch
  • Setup rviz :
  • set fixed frame to camera_rgb_optical_frame.
  • add an InteractiveMarker and set topic to (server_topic parameter)/update.
  • add a Marker and set topic to vizualization_topic parameter.
  • add the Rviz interface panel.
  • add a PointCloud2 and set topic to /camera/depth_registered/points.

ASIFT matching (no vizualisation)

  • Run :
  • roslaunch openni_launch openni.launch
  • asift_matching : roslaunch asift_matching asift_match.launch

ASIFT matching & Rviz interface

  • Run :
  • roslaunch openni_launch openni.launch
  • roslaunch rviz_interface interface_matching.launch
  • Setup rviz :
  • set fixed frame to camera_rgb_optical_frame.
  • add an InteractiveMarker and set topic to (server_topic parameter)/update.
  • add a Marker and set topic to vizualization_topic parameter.
  • add the Rviz interface panel.
  • add a PointCloud2 and set topic to /camera/depth_registered/points.

Parameters

  • The topic names are defined as parameters in the launchfiles.

  • rviz_interface parameters are defined as parameters in the rviz_interface launchfiles : Interface objects are defined by their names and types (0-Default, 1-Button, 2-2D object, 3-3D object) (see StateSpace.msg for more details).

  • asift_matching parameters are defined as parameters in the asift_matching launchfiles : -> References are defined by the path to the references files can be a txt file (faster - see texte reference generation in asift_matching section) or a series of images files (slower - currently not working). -> Tracked object is defined by the object name (must be one of the interface object to work properly with rviz_interface). -> Number of tilts is defined by an integer value (1-no tilt, 7-recommended value). Higher value means slower process. -> Standaed deviation filter coefficient is defined by a real value. Keep (1-68%/2-95%/3-99%) of the points statistically. 0: no filtering (default).