Currently, segmentation.cpp in hlpr_segmentation can clip the input point cloud to a bounding box with a specified size. However, this bounding box is specified in the camera reference frame, which can result in unintuitive behavior if the camera is not aligned with the robot base frame.
The clipping parameters should be specified in the robot's torso frame, neck frame, or arm frame. To perform clipping, we'll have to transform points into that frame using a passthrough filter (http://pointclouds.org/documentation/tutorials/passthrough.php#passthrough).
Currently, segmentation.cpp in hlpr_segmentation can clip the input point cloud to a bounding box with a specified size. However, this bounding box is specified in the camera reference frame, which can result in unintuitive behavior if the camera is not aligned with the robot base frame.
The clipping parameters should be specified in the robot's torso frame, neck frame, or arm frame. To perform clipping, we'll have to transform points into that frame using a passthrough filter (http://pointclouds.org/documentation/tutorials/passthrough.php#passthrough).