Only released in EOL distros:  

roboearth: ar_bounding_box | re_2dmap_extractor | re_comm | re_comm_core | re_kinect_object_detector | re_msgs | re_object_detector_gui | re_object_recorder | re_ontology | re_srvs | re_vision

Package Summary

This package implements a detection method for models created by re_object_recorder.

This package is part of RoboEarth (EU FP7, grant 248942).

What does re_kinect_object_detector do?

Once you recorded a model using re_object_recorder, you can use re_kinect_object_detector to detect it again in a point cloud.

See the tutorial roboearth/Tutorials/Using recorded models for object detection for a usage description.

Interface

re_kinect

Main node for recognizing objects with the Kinect camera.

Subscribed Topics

camera/depth_registered/points (sensor_msgs/PointCloud2)
  • Input point cloud to in which to look for known objects (usually straight from [openni_node]).
re_kinect/model_path (std_msgs/String)
  • Signals re_kinect to load the model stored at the given path.

Published Topics

re_kinect/feature_pcl (sensor_msgs/PointCloud2)
  • The matched model aspect as point cloud.
re_kinect/detection_results (re_kinect_object_detector/DetectionResult)
  • Announces detected objects along with the input image and detected object names.

Provided tf Transforms

/camera_rgb_optical_frameOBJECT_CLASS.OBJECT_NAME
  • Transformation from the camera frame to the object frame.

Wiki: re_kinect_object_detector (last edited 2011-11-16 14:36:28 by DanielDiMarco)