[Documentation] [TitleIndex] [WordIndex

Only released in EOL distros:  

Description

Voxelized Shape and Color Histograms (VOSCH) is an efficient rotation-invariant descriptor extracted from color voxel data. It is obtained as a 137-dimensional histogram, where each bin is represented by local correlations of surface normal types or RGB color values.

Thus the descriptor enables object classification and identification considering both shape and color characteristics. It can be used with any type of classifier, e.g., Support Vector Machine (SVM) or Linear Subspace Method (LSM).

vosch.png

Install

git clone git://github.com/kanezaki/isi-ros-pkg.git
git clone http://code.in.tum.de/git/mapping.git

Usage

$ rosrun vosch example_vosch `rospack find vosch`/data/sample_cone_green.pcd

This executable reads color point cloud "sample_cone_green.pcd" and writes its VOSCH descriptor into "vosch.pcd" in the current directory.

If you have pcl_visualization, you can see the histogram by:

$ rosrun pcl_visualization pcd_viewer vosch.pcd

vosch_cone.png vosch_torus.png

The Code

The following code (test/example_vosch.cpp) snippet will estimate a set of VOSCH features for all the points in the input dataset.

Could not fetch external code from 'http://code.in.tum.de/indefero/index.php//p/mapping/source/file/master/vosch/test/example_vosch.cpp':

If you use the following two lines instead of the corresponding lines in above example code, you can get a set of VOSCH feature vectors extracted from the subdivisions of 10x10x10 voxel grid with 5x5x5 overlapping respectively:

std::vector< std::vector<float> > vosch;
extractVOSCH( grid, cloud, cloud_downsampled, vosch, 127, 127, 127, voxel_size, 10, 5, 5, 5 );


2019-10-12 13:16