VVV09 visual attention

From Wiki for iCub and Friends
Revision as of 16:36, 29 July 2009 by VVV09 (talk | contribs)
(diff) ← Older revision | Latest revision (diff) | Newer revision → (diff)
Jump to: navigation, search

This group is composed by people with an interest in Computer Vision, please add your name to the list if you're one of us :-). Please feel free to edit this page at will.

Members: Katrin, Matteo, Nicola, Federico_T., Harold, Fabrizio, Jakob, Giovanni.

Some of us are collaborating with the VVV09_Manipulation_Group with the intent of providing them with usable cuboid trackers.


ARToolkit-based cuboid tracking

Katrin is currently working on this. There is a bug in the ARToolKitTracker, in the visualization part I have done a temporaly fix. So the Tracker is runing. If you are interest in using it ask me until the bug is fixt for real.

Colour-based cuboid tracking

Matteo is currently working on this. The goal of this activity is to create a fully YARPed C++ cuboid tracker that provide information to the grasping modules.

Full bullet 2.png make a simulator to generate synthetic image sequences of moving cubes


Full bullet 2.png gather some realistic data

RealLegoCube1.jpg RealLegoCube2.jpg

Should you need this set of images, please ask Matteo: the set is a bit bulky, it won't be uploaded somewhere to avoid stressing the network.

Empty bullet.png make the tracker work in Matlab with the synthetic image sequences

SyntheticCubeTracked1.jpg SyntheticCubeTracked55.jpg

This one is starting to work, at least for a few feames. Please have a look at this video.

Empty bullet.png make the tracker work in Matlab with real image sequences of moving cubes 
Empty bullet.png make the tracker work in C++ with real image sequences of moving cubes 
Empty bullet.png design and implement a system for the initialization of the tracker 

Colour-based Ball (blob) tracking

Matteo is debugging the pf3dTracker you find in the iCub repository. Should you need an easy to use tracker, please try it out and complain to Matteo when it doesn't work :-). You can find information on how to run the ball tracker here.

Calculate object orientation and position from perspective information

I have tried to calculate the position and the orientation of an object using perspective information extracted from the image.

- extract edges of the objects

- group the edges and find parallels to calculate vanishing points and vanishing line

- use the vanishing line and the vanishing points to calculate the camera position relative to the object


Edges.jpg Lines.jpg Vanishing line.jpg

I have extracted the edges of the objects and calculate the vanishing points and the vanishing line using the edges.


Here you can find a set of images illustrating possible operating conditions for the grasp demo.