The MOPED framework: Object recognition and pose estimation for manipulation

This paper describes the MOPED framework for object recognition via cameras and describes what inputs it needs, how features get matched between database and the extracted image from the cameras, complexity of using the algorithms for these purposes and scalability of the MOPED. There is also data on the recognition accuracy of the MOPED.

This paper could be useful to get an idea on how object recognition works and what is needed for it.