robokudo.annotators.region_filter

Point cloud filtering based on semantic regions.

This module provides an annotator for:

  • Filtering point clouds using semantic map regions

  • Supporting regions in world and local frames

  • Creating region hypotheses with poses

  • Visualizing filtered regions and clouds

The module uses:

  • Semantic maps to define regions of interest

  • Frame transformations for region alignment

  • Oriented bounding boxes for point filtering

Note

Regions can be defined in either world frame or local frames.

Classes

RegionFilter

Point cloud filtering using semantic map regions.

Module Contents

class robokudo.annotators.region_filter.RegionFilter(name='RegionFilter', descriptor=Descriptor())

Bases: robokudo.annotators.core.ThreadedAnnotator

Point cloud filtering using semantic map regions.

The RegionFilter can be used to filter point clouds based on a environment model based on different regions. These regions are collected in a ‘SemanticMap’ which has one ‘SemanticMapEntry’ per region of interest. Semantics to these regions are linked by referencing well-known names from your URDF and/or knowledge base.

Note

Regions are defined in a SemanticMap with SemanticMapEntry objects.

class Descriptor

Bases: robokudo.annotators.core.BaseAnnotator.Descriptor

Configuration descriptor for region filtering.

class Parameters

Parameters for configuring region filtering.

Frame parameters:

Variables:

world_frame_name – Name of the world coordinate frame, defaults to “map”

Semantic map parameters:

Variables:

  • semantic_map_ros_package – ROS package containing semantic map, defaults to “robokudo”

  • semantic_map_name – Name of semantic map module, defaults to “semantic_map_iai_kitchen”

Region selection:

Variables:

active_region – Name of active region to filter, empty for all regions

world_frame_name = 'map'
semantic_map_ros_package = 'robokudo'
semantic_map_name = 'semantic_map_iai_kitchen'
active_region = ''
parameters
world_frame_name
semantic_map = None
active_region
load_semantic_map() → None

Load semantic map from configured package and module.

Uses ModuleLoader to dynamically load the semantic map module.

compute()

Filter point cloud using semantic map regions.

The method:

  • Loads point cloud and query from CAS

  • Updates semantic map and active regions

  • For each active region: * Transforms region to camera frame * Creates oriented bounding box * Filters points within box * Creates region hypothesis with pose

  • Updates CAS with filtered cloud

  • Creates visualization markers

Returns:

SUCCESS after processing

Return type:

py_trees.Status

Raises:

Exception – If queried region not found in map