visual-tracking-control
hand-tracking

A visual end-effector tracker using a 3D model-aided particle filter. More...

A visual end-effector tracker using a 3D model-aided particle filter.

Version: 0.6.0.0

Author
Claudio Fantacci claud.nosp@m.io.f.nosp@m.antac.nosp@m.ci@i.nosp@m.it.it

Description

This module run a recursive Bayesian filter, a particle filter, to track the robot's end-effector.

For each particle filter recursion

1) we render, for each particle, an image of the 3D mesh model of the end-effector as it would appear from the robot’s viewpoints (likewise in augmented reality contexts);

2) we then use this state representation to directly estimate the 6D pose (position and orientation) of the end-effector in the robot operative space using 2D image descriptors.

In particular, we use HOG to compare the rendered images with the robot’s camera images and, as a result, the particles which are more likely to represent the end-effector will have higher weight.

Parameters

Input Ports

Output Ports

Services