seg2cloud

Module to capture pointclouds from segmented blobs in 2D or 2.5D images

Description

This module combines disparity and segmentation information in order to retrieve 3D clouds of desired objects. The objects can be chosen from a user defined polygon, from a segmentation algorithm, or from 2D (in color) or 3D (in depth) flood filling. When using segmentation modules (2D or 3D), seeding can be performed automatically.

Dependencies

• YARP
• iCub
• icub-contrib-common
• OpenCV
• SFM

Reconstruction Options

This modules interfaces between segmentation modules and SFM in order to provide the 3D reconstruction of an object delimited by a contour in the image. In order to obtain this contour, 4 modes are provided:

• seg: 2D segmentation (requires external segmentation module). When this option is called, the contour is provided by an external segmentation module.
• flood: 2D color flood. The object's contour is obtained by performing a 2D color flood, based on the thresholding parameter color_distance.
• flood3D 3D depth flood. The contour is determined by a flood operation performed on the depth image, controlled by the thresholding parameter space_distance.
• polygon User selected polygon contour. The contour is provided by the user as a contour by selecting the vertices on a connected yarpview window.

The results obtained with each of these options can be observed below.

In the three first options above (seg, flood and flood3d) the initial seed to perform segmentation can be provided in 3 different ways:

• If parameter seedAuto is set to true, the seed will be obtained automatically, by reading from a tracker (2D options seg and flood) or selecting the closest blob (3D option flood3d).
• If parameter seedAuto is set to false, the seed can be coveyed in 2 further ways:
  • By clicking on the a connected yarpview window with output
  • As explicit u v coordinates given after the segmentation command. For example seg 102 175 .

Documentation

Online documentation is available here: http://robotology.github.com/segmentation-to-pointcloud.

Images of reconstruction results:

###seg (2D segmentation provided by lbpExptract)

###flood (2D color flood)

###flood3D (3D spatial flood)

###polygon (Selected polygon segmentation)

License

Material included here is Copyright of iCub Facility - Istituto Italiano di Tecnologia and is released under the terms of the GPL v2.0 or later. See the file LICENSE for details.