Abstract: This technique is a feature-based approach to detection that uses machine learning to select the best features. These features encapsulate intensity-distribution domain data about a region. This kind of data is very flexible to minor changes in images, so it is well suited to the detection of robots which often vibrate, move, and change shape. To give a complete cover, we split the detector into 5 simultaneous cases, shown in the image, and then combine the results with a voting technique to select the final match.

Researchers: James E. Young, Ehud Sharlin, Jeffery E. Boyd

• Publications:
  1. J. E. Young, E. Sharlin, and J. E. Boyd, “Implementing Bubblegrams: The Use of Haar-Like Features for Human-Robot Interaction”, IEEE Conference on Automation Science and Engineering, CASE 2006, Shanghai, China, October 7 – 10, 2006.
  
  

Video:

Difficult case

Detector cases

AIBO