Only released in EOL distros:   fuerte

What does re_vision do?

The re_vision package contains the ObjectDetector node, that allows to recognize known objects in images. The data returned are the 3D pose of the object and the location in the image of some pixels that belong to the object.

There is also an ObjectDetectorDummy node that allows to recognize a predefined local object in the gazebo simulator.

Invocation

$ rosrun re_vision ObjectDetector [options] [camera_info:=/your/camera_info]
$ rosrun re_vision ObjectDetectorDummy [options] [camera_info:=/your/camera_info]
$ roslaunch re_vision gazebo.launch

The first line runs the ObjectDetector node, which must be notified the object models to load via topic. The ObjectDetectorDummy behaves the same as the former, but it also loads a predefined model, available for the gazebo simulator (name "cabinet1"). The easiest way to run ObjectDetectorDummy is with the gazebo.launch file, because you do not have to take care of the location of the predefined model. Both nodes need that parameters of the camera that provides the images. These are obtained from the /camera_info topic (see below). You may want to remap this topic in the gazebo.launch file or in the command line.

Options:

• -d: debug flag. If given, debug images of the recognition process are saved in ./debug/debug_[timestamp].

• -v: visualization flag. If given, detection images are sent by the topic /re_vision/detector_visualization.

• -c: predefined_camera: load parameters of a predefined camera. This option is used for debugging purposes only. The camera models and their parameters are defined in re_vision/src/CameraBridgeFactory.cpp.

Interface

ROS Topics

Subscribed Topics

/camera_info (sensor_msgs/CameraInfo)

• This topic is used to retrieve the intrinsic parameters and the distortion model of the camera that will provide the images. The information is read only once; after that, the topic is not listened any longer. You may want to remap this topic when running the ObjectDetector.

/re_vslam/new_model (std_msgs/String)

• When a new object model has been downloaded from the RoboEarth database, ObjectDetector is notified with this topic that the new model is available. After getting this topic, the model is learnt, so that it can be searched for in the following given images.

Published Topics

/re_vision/detector_visualization (sensor_msgs/Image)

• If the -v option is given, ObjectDetector sends images of the recognition process by this topic. These images can be later visualized with image_view.

ROS Services

Services

/re_vision/search_for (re_vision/SearchFor)

• Searchs the image for the given objects. Only those objects whose name are given are searched for. If an object is recognized in the image, up to MaxPointsPerObject points from this are returned. If MaxPointsPerObject is negative, all of them are returned. If the i-th object given is not detected, the corresponding arrays in Detections will be empty. The axes in the camera reference system are oriented as in the robot: X looks forward, Y points left and Z goes up.