Cloud Transformation and Processing

[André]

Hi guys,

I would like to perform a cloud processing in the raw cloud the Kinect receives from the OpenNI.
Which class get the raw cloud and pass it for the slam processing?

I was editing the main file in rtabmap/tools/CameraRGBD, but I commented the whole class, compiled and nothing seemed to change in the application.

Thanks a lot, cheers!

[matlabbe]

What kind of cloud processing do you want to do? Generation of the point clouds are done in the GUI only for vizualization purpose. The data kept for actual SLAM are raw RGB and depth 2D images. For example, when we show a point cloud in the 3D Map view, the point cloud is generated using this method MainWindow::createCloud(). For the actual method creating a point cloud from RGB and depth images acquired from a CameraRGBD object, see util3d::cloudFromDepthRGB().

Cloud processing on these created clouds won't affect the SLAM system.

rtabmap/tools/CameraRGBD is a simple application called "rtabmap-rgbd_camera" to show RGB and depth images as well as creating a point cloud to show.

[André]

Hi, Matlabee,

I would like to detect the floor and paint in red before joining the clouds by matching the keypoints. Do I have to modify util3d.cpp to do that? Thanks =)

[matlabbe]

I would do that in guilib/src/MainWindow.cpp in the method MainWindow::createCloud(). There is some post processing on the point cloud here, you can put your filter with them. For plane segmentation, I would refer to the planar segmentation tutorial of PCL documentation.

[Leo]

The util3d :: cloudFromDepthRGB () class, served to what we were trying to do in standlone version but would like to know which class is similar or equal to util3d :: cloudFromDepthRGB (). in the version for ROS so that we can test in ROS version as well ?

Thanks

[matlabbe]

The rtabmap_ros/point_cloud_xyzrgb nodelet does something similar. The code is here. Actually, it is a wrapper of the util3d::cloudFromDepthRGB() method.