ROS Resources: Documentation | Support | Discussion Forum | Index | Service Status | ros @ Robotics Stack Exchange
Ask Your Question

Which is the easiest way to do Object Detection

asked 2013-09-12 07:48:34 -0500

ctguell gravatar image

Right now ive been searching the web and got that th best packages are roboearth and tabletop_object_detector, but I have to do the specific of recognizing doors and people, so im no sure this are the best stacks to do the recognition of this two clases. Ive found the pr2_doors and openni_tracker but would really appreciate some help from someone who actually accomplished to do recognition of doors or people.

edit retag flag offensive close merge delete

1 Answer

Sort by ยป oldest newest most voted

answered 2013-09-12 20:30:54 -0500

moritz gravatar image

Both methods won't work for doors or people. The tabletop object detector is based on the assumption that all objects stand on a flat horizontal surface that can be used to segment the items (the tabletop). This is not the case for doors and hardly for people. The RoboEarth object recognizer is based on local image features (SURF), which means that it only works with non-deformable textured objects.

I'm not aware of any out-of-the box package for your use case, but there are components. openni_tracker is able to detect persons, opencv should provide methods for recognizing faces, and building a door detector using PCL should be fairly easy. As always, it depends on your setting, i.e. which sensors, which kinds of doors, open/closed, ...

edit flag offensive delete link more

Question Tools



Asked: 2013-09-12 07:48:34 -0500

Seen: 960 times

Last updated: Sep 12 '13